CN110039173A - Radium-shine system of processing and method for beam jitter and scraping - Google Patents

Abstract

A kind of radium-shine system of processing includes the first positioning system for assigning the first relative motion of the beam path along the beam trajectory relative to workpiece, for determining the processor of the second relative motion of the beam path along multiple shake rows, for assigning the second positioning system of second relative motion, and the radium-shine source of sending radium-shine light beam pulse.The system can compensate for the variation in process velocity to retain shake row at a predetermined angle.For example, the shake row can ignore process velocity and remain perpendicular to the beam trajectory.The process velocity can be adjusted and complete groove to process to integral jitter row.The quantity of the jitter points in every a line can be selected based on the groove width.Fluence can be normalized by the adjustment to process velocity and the variation of groove width.

Description

Radium-shine system of processing and method for beam jitter and scraping

The application is October 24 2011 applying date, application number 201610090031.X, invention and created name " for light Beam shake and the anchor of scraping penetrate system of processing and method " divisional application.

Technical field

The present invention relates to the radium-shine processing of dielectric or other materials.

Background technique

The radium-shine processing of dielectric and conductive material is commonly used to the fine feature in ablation electronic building brick.For example, can be with radium Processing chip package pedestal is penetrated so that signal is routed to ball grid array or similar encapsulation from the semiconductor dies.Radium-shine machining feature It may include signal trajectory, ground wire and micropore (for being connected to the signal trajectory between encapsulated layer).

Radium-shine direct ablation (LDA) merges signal and ground wire to reduce in the quantity of chip package internal layer in single layer When strict control signal impedance.Such mode may need lesser characteristic size and space (e.g., to arrive for about 10 microns (μm) About 25 μm) and each encapsulation long path length (e.g., about 5 meters (m) to about 10 meters).In order to economically construct chip It encapsulates, speed locating for the such feature of ablation may be quite high (e.g., from about 1 meter per second (m/s) to about 10m/s).It is processable Certain encapsulation, for example, meeting customer's quantum of output target to about 5s using about 0.5 second (s).

Another useful properties of chip package may be to provide cross track using the change in depth controlled.Example Such as, ground wire branch can occur at several through the pattern.In each branch's infall, required be less than greatly can be used About +/- 10% change in depth comes these tracks of ablation.Generally, if by two grooves of ablation, the ablation at a point The double exposure of light beam by create about 100% change in depth.

Another useful properties of chip package may be at the different piece of the encapsulation provide track changeable width with It controls impedance or the pad for interlayer connection via hole is provided.To there is the track width control of rupture reduce or minimum to provide To the High-speed machining of the backbone mark.

It is equally possible useful, it is processed with using the high speed for the reduction or minimum time that change this feature characteristic The feature of arbitrary size and shape.For example, feature may include the micropore with multiple diameter and/or side wall taper, square or Rectangular pads are aligned benchmark, and/or alphabetical numerical digit mark.Traditionally, in order to process the feature such as micropore, optics has been designed System is used to provide the forming light distribution (e.g., flat top beam) or pure Gaussian beam of variable-diameter.Add when change is radium-shine When work dot characteristics, these photosystems may have significant time delay (e.g., arriving about 10s for about 10 milliseconds (ms)).

Other problems are related to a kind of machine for meeting above-mentioned machined parameters is established.For example, track is in the encapsulation process Since routing needs that direction may be changed.When with High-speed machining track, the variation at the trajectory angle may be needed very short High beam position at markers accelerates.Radium-shine processing can easily exceed the dynamic limit of the beam positioner, for example, when with High speed for high yield output is (when e.g., about 1m/s to about 10m/s) is run.

Such acceleration and/or speed may be difficult to reach in traditional radium-shine processing machine, have relied in Light beam location technology such as merges the linear of mirror galvanometer beam deflector (being here referred to as " galvanometer " or " galvanometer mirror ") Stage, with can not for the type processing (e.g., based on sequence) from about 1 microsecond (μ sec) to about 100 μ sec when Static state (or slowly varying in) light beam pretreatment optics responded in mark is together.

The practical ablation may be also intended to the factor considered.Laser pulse with high-peak power can be used for ablation The dielectric material and minimize as melt, cracking and pedestal damage hot side effect.For example, having with about 5 megahertzs (MHz) Repetitive rate to about 100MHz arrives the ultrafast radium-shine of the pulse width in the range between about 50ps at about 20 picoseconds (ps) The material with high-peak power can be processed and improve significant pulse and overlap to avoid pulse spacing effect.Now, optical fiber It is radium-shine generally to provide the pulse width for being more than about 500 kHz (kHz) repetitive rates in nanosecond region.Generally, right In given processing situation (ablation depth and width), " dosage " (power/speed) for the rapidoprint should be constant.So And at low velocity, the power of the application may become so low that the peak pulse power may be not enough to the ablation material Material is without causing hot side effect (e.g., melt and carbonize).

Beam positioner design may deflect the machining beams using galvanometer.The light of the machining beams at workpiece Strong distribution may be (the simple focusing for Gaussian beam) of Gauss, or for pretreated by fixed beam shaping optics device The forming light distribution (e.g., flat-top distribution) of light beam.

Summary of the invention

In one example, for shake the method for radium-shine light beam form along beam trajectory and the process velocity of variation without The groove with groove width needed for one or more in workpiece closed.This method includes assigning along relative to the workpiece First relative motion in the radium-shine light beam path of the beam trajectory on surface, and determine the radium-shine light beam road along multiple shake rows Second relative motion of diameter.Second relative motion is first opposite at this with the predetermined angular superposition relative to the beam trajectory On movement.The determination of second relative motion includes for for multiple shake row, each to retain the processing of the predetermined angular The compensation changed in speed.This method further comprises second relative motion in the imparting radium-shine light beam path, and along this Multiple radium-shine light beam pulses the workpiece is dealt at multiple positions of multiple shake rows to be used to the defined side of the predetermined angular To widening groove.In some instances, the predetermined angular is perpendicular to the beam trajectory.In addition, or in another example, issuing should Multiple radium-shine light beam pulses include being issued with constant speed, this method further comprise selectively adjust the process velocity in order to Integral jitter row is processed to complete the groove.

In another example, radium-shine system of processing includes the first positioning system, for assigning along relative to the workpiece table First relative motion and one or more processors in the radium-shine light beam path of the beam trajectory in face, for determining along multiple Shake second relative motion in the radium-shine light beam path of row.By second relative motion relative to the predetermined of the beam trajectory Angle superposition is on first relative motion.The determination of second relative motion includes for going each for multiple shake Retain the compensation changed in the process velocity along the beam trajectory of the predetermined angular.The system also includes the second positioning system System, for assign the radium-shine light beam path the second relative motion and radium-shine source, for along it is multiple shake row it is multiple Multiple radium-shine light beam pulses the workpiece is dealt at point position to be used to widen groove with the defined direction of the predetermined angular.

In another example, groove is created in workpiece for shaking the method for radium-shine light beam.This method includes assigning Along first relative motion in the radium-shine light beam path for the beam trajectory for defining the trench length relative to the workpiece surface, and Assign second relative motion in the radium-shine light beam path along multiple shake rows.By the second relative motion superposition this first To widen the groove on relative motion.The variable-width of the groove.This method further comprise select jitter points quantity with Including in each of multiple shake row.The selection reduces the time quantum for processing each shake row, each The quantity of jitter points in a shake row is based on the groove width for corresponding to each shake row.This method further comprises right Multiple radium-shine light beam pulses are dealt into the workpiece at multiple positions of jitter points of the Ying Yu in multiple shake row in each.

In another example, the method in the two-dimentional scraping region for radium-shine processing on workpiece has used radium-shine direct burning Erosion system.This method includes that radium-shine position grid is generated in the scraping region.Between radium-shine position in the grid Interval is at least partially based on the required overlapping of radium spot size and adjacent radium-shine point.This method further comprises corresponding to along light The individual channel in the radium-shine light beam path of beam trajectory is by the grid dividing at multiple bands.Each band includes relative to the light beam rail Mark along jitter direction multiple shake rows.This method also includes assigning being somebody's turn to do continually by multiple bands along the beam trajectory First relative motion in radium-shine light beam path assigns radium-shine light beam path of the jitter direction of each shake row along this Second relative motion, and along the radium-shine light beam path to radium-shine position in the scraping region by multiple radium-shine light beam arteries and veins Punching is sent to the workpiece.In some instances, this method further comprise to the shake row each data filter to tilt Make along radium-shine light distribution of the side wall of the adjacent belts of the grid come the overlapping forming between the access to the radium-shine light beam It obtains and controls change in depth in the scraping region.The radium of each of multiple radium-shine light beam pulse is selected based on the filtering data Exit point intensity.In addition, or in another example, this method includes moving on to the net from the first band of the grid when the light beam path The jitter direction is overturn when the second band of lattice.The jitter direction can be based on the direction of the beam trajectory.

From the point of view of the following specifically daylight of the preferred embodiment continued together with the label to the attached drawing, additional side Face and advantage will be apparent.

Detailed description of the invention

Fig. 1 depicts the schematic diagram of the grid of processing stand, can by jitter motion light beam or fixed grating pattern come It generates；

Fig. 2 show graphically according to an example be used as opposite mesh width (being standardized by spot diameter) and in the grid it is wide The EffectiveWidth of the function of the quantity of jitter points on degree changes；

Fig. 3 describes the lattice for two cross features for having different mesh-densities according to an example with figure；

Fig. 4 describes the modelling of the cross feature after amplitude normalizes (not merging) according to an example with figure Fluence；

Fig. 5 describes the merging fluence according to the cross feature of an example with figure；

Fig. 6 describes the rasterizer grid (right side) and fluence for the circular pad with trimmer according to an example with figure It is distributed on (left side)；

Fig. 7 describes the influence to standardize according to the fluence on the groove of varying width of an example with figure；

Fig. 8 describes the exemplary process calibration test matrix according to an example；

Fig. 9 describes the examples material table according to an example；

Figure 10 is according to example description for ordering the block diagram for simplifying processing and data process structure of the AOD；

Figure 10 A is according to an example include AOD subsystem and galvanometer subsystem for shaking radium-shine light beam system Block diagram；

Figure 10 B is the block diagram for being used to the system of beam-shaping according to an example；

Figure 10 C is that the block diagram of the system of inclination machining beams is provided according to an example；

Figure 11 is to be indicated according to an example with the block diagram of the FPGA AOD control data process implemented；

Figure 12 is the timing diagram that the example sync of DSP and FPGA are described according to an example；

Figure 13 is the schematic diagram that the processing situation using landslide is described according to an example；

Figure 14 is described with figure according to the example of example landslide parameter scaling；

Figure 15 describes the example XY light-beam position of dither operation according to the example with figure；

Figure 16 describes example X and Y the light-beam position vs. time according to example dither operation in Figure 15 with figure；

Figure 17 describes the example light-beam position of example dither operation in Figure 15 and 16 and the table of machined parameters；

Figure 18 describes the scanning field distortion pattern in sample instance with figure, and wherein the LDA system includes F/18 100mm scans field lens；

Figure 19 describes vs.X and Y positions X and Y scan field distortion error according to sample instance with figure；

Figure 20 describes the X and Y scan field distortion scale factor and rotation error according to the sample instance in Figure 19 with figure Vs.X and Y position；

Figure 21 describes the example local position geometry amendment according to the spectra features of example in Figure 19 and 20 with figure Distortion；

Figure 22 describes the third level section subsystem according to an example with figure；

Figure 23 describes the AOD calibrating pattern according to an example with figure；

Figure 24 describes the AOD collimation angle according to certain example with figure；

Figure 25 depicts the block diagram according to an example for the signal stream of power control；

Figure 26 describes the example power controlling curve according to an example with figure；

Figure 27 describes the example ch1AOD power linear curve set according to an example with figure；

Figure 28 describes the example plot for ch0AOD according to an example；

Figure 29 is described according to the figure of the example A OD efficiency curve of certain example；

Figure 30 is described according to the figure of the example A OD efficiency gain of certain example；

Figure 31 is the block diagram for describing the calibration of power data process according to an example；

Figure 32 describes the third level filter in response to velocity variations according to an example with figure；

Figure 33 describes the Monte-Carlo AOD Transient according to an example with figure；

Figure 34 describes the example rate limitation using Taod=Tcmd=1 μ sec according to an example with figure；

Figure 35 describes the AOD stroke range vs. characteristic width according to an example with figure；

Figure 36 describes the radium-shine power limit relevant to process velocity according to an example with figure；

Figure 37 is described with figure and is distorted according to the example points of certain example；

Figure 38 describes the shake row for being vertically aligned at nominal beam track according to an example with figure；

Figure 39 describes the example of the tiling grating processing according to an example with figure；

Figure 40 describes the dot matrixed error for being attributed to the wide line scraping according to certain example with figure；

Figure 41 describes the scraping region for being divided into network band according to an example with figure；

Figure 42 is described with figure according to the unfiltered grids of the scraping jitter points of an example and corresponding scraping jitter points Filter grid；

Figure 43 describes the fluence for corresponding to first three scraping light beam of the band in Figure 41 according to an example with figure Distribution；

Figure 44 describes the scraping example according to an example with figure；

Figure 45 describes the light beam order during the scraping according to an example with figure；

Figure 46 depicts the figure of the material removing speed vs. scraping row length and width according to an example；

Figure 47 describes the side view of the scraping row geometry according to an example with figure；

Figure 48 is the block diagram generated for the AOD order of vector processing described according to an example；

Figure 49 is the block diagram generated for the AOD order of grating processing described according to an example；

Figure 50 is the block diagram for describing to be generated according to the AOD order for shaving action of an example；

Figure 51 is the block diagram for describing to calibrate data process 5100 according to the galvanometer and AOD of an example；

Figure 52 describes the shake row dose control according to an example with figure.

Specific embodiment

Be incorporated herein present invention illustrates use acoustics-optical deflector (AOD) as the LDA system of light beam positioning device The details of the embodiment of system.When this exemplary case disclosed herein is related to AOD, it is possible to use electro-optical deflector (EOD).In some instances, for example, EOD is the appropriate replacement that some or all of AOD are directed toward (deflection) function.

In some instances, the geometry (e.g., width and depth) of machining feature can be appointed as by the user high Grade, and Machining Instruction is then converted by apparatus control software.The details of radium-shine power and the dither operation in some instances It is automatically to prevent to be difficult to manual machines installation process bear, fallibility.For example, can provide creation for user has nominally The simple processing of the geometric characteristic of size.This feature may include the groove with target width and depth, or have target The pad of diameter and depth.The user can directly input these geometric parameters, and the system generates this feature institute by being generated as The correct machined parameters (e.g., beam velocity shakes width, radium-shine power) needed are responded.Because certain LDA machines can Operate in arbitrary beam speed (in order to maximize given quantum of output beam positioner and radium-shine power constraint), can automatically for speed Degree is to adjust the machined parameters.This avoid automatically force the user point out inferior grade details (e.g., radium-shine power, shake width, Jitter points, speed), the risk with attached operator's mistake, and the system is allowed automatically to maximize quantum of output.

Certain examples provide the jitter points selection of optimization.The example is by determining that shake table size has at least to cover The required jitter range of count jitter point maximizes quantum of output.One reason of the minimized jitter points quantity is because each A jitter points have used certain renewal time Tdither (e.g., according to the update of about 1 μ s of sample instance discussed below speed Degree).In place of Npts is the jitter points quantity of every row, Tdither*Npts=is used for the time quantum of every row.Thus, minimizing should Jitter points quantity allows the system to carry out machining feature with highest possible speed and be retained in the necessary overlapping between shake row, has For create mean fluence distribution, and thus this feature average ablation.For example, this is trembled when shaking to form feature using AOD Dynamic point remains minimum overlap to minimize fluence variation.The overlapping can be influenced for widening the jitter points quantity of groove.? Optimize in certain examples the selection of the jitter points with and meanwhile provide high process velocity and sufficient point overlaps.

Certain examples provide fluence standardization for wide arc machining.Such example is remained in the width to arc radius The quality (e.g., uniform trench depth) of the ablation circular arc feature in the case where the ratio of degree is relatively large.This sets for circuit design Meter person provides more freedom being routed to circular arc than in other feasible more fastening areas.For example, when shake is to process When forming the wide groove of circular arc component, letter of the used fluence as the radius at the center of arc on the width groove Number is to be changed.The fluence is standardized in some instances and processes the circular arc to use uniform depth.

Certain examples provide the coordination calibration of AOD and galvanometer positioning system.Other positioning systems can also be used, such as make The behaviour of the shake on the entire scanning field is remained with the system that the quick manipulation mirror (FSM) of different positioning systems coordinates calibration Precision in work.The AOD light beam positioning subsystem can suffer from being attributed to the scanning lens, galvanometer mirror and AOD subsystem The scanning field distortion of embodiment.For example, scanning lens generally has the local geometric distortion with scanning field change in location.It is logical The function that local amendment is used as to scanning field position is crossed, which has been retained in be used to provide can receive deeply in infall Spend the grade of variation.In addition, or in other instances, the calibration of the AOD light beam positioning subsystem is coordinated in the galvanometer subsystem Correct embodiment of the calibration of system for the positioning of third level light beam.It as following discussion, is shaking, grating and scrape mode Period uses such calibration.This can be positioned in conjunction with the third level light beam to carry out, and the AOD subsystem is ordered to keep track Center line (during vector shake or scrape mode) or grating midpoint (during grating mode).

Certain examples provide the grating pattern of overlapping.No matter when quantum of output can be optimized by spectra features, due to The AOD raster manipulation occurs positioning very high speed than possible service stage or galvanometer light beam.However, the AOD Confining field size limits in one step can be by the feature sizes of AOD grating.By overlapping grating pattern, when heavy at this Repeatedly the infall in region keeps correctly deep-controlled, can be the pattern generation high-quality grating pattern beyond the AOD size, Thus maximize quantum of output.In some instances, which is used to the grating two dimensional character within the scope of the AOD, without Any movement from the galvanometer subsystem.It is multiple independent by overlapping at different (and fixed) galvanometer coordinates Grating pattern also can carry out grating to the pattern greater than AOD.This mode may be more efficiently (thus with greater need for) Processing method, for the General Two-Dimensional scraping than using movement inspection trajectory mark.As employed herein, two-dimentional (2D) is related to adding The region work 2D (e.g., in X and Y-direction), but also include the deep-controlled radium-shine processing using three-dimensional (3D).Deep-controlled packet It includes, for example, the quantity of material for being removed with Z-direction or being shaped to the machining feature.

Certain examples provide the AOD efficiency optimized compared with range.The production realized when the machining feature in LDA system Output to provide the required machined parameters to can get radium-shine power proportional.Optimize the radium-shine power thus optimizes quantum of output. By optimizing the AOD efficiency as required AOD opereating specification function (when keeping accurate AOD power on the field the AOD of the operation When linear), it can reach such power optimization.Quantum of output is maximized when machining feature needs the relatively small field AOD, and working as needs It wants still to provide the performance for maximizing the AOD range when larger feature.For example, when using AOD to shake or formed two-dimensional grating figure When case, with the range of deflection of growth, which is reduced.Thus, it may be desirable to be that can be linearized in opereating specification at this Optical output power at the AOD is so that AOD operation is predictable and consistent.In some instances, linearisation is provided Power control, has used the high light efficiency of the reduction AOD range of deflection in certain operations, and allows for other operations Obtainable larger range (there is reduced optical efficiency).

Certain examples provide speed-optimization using third level filtering.The third level filtering mode, merges conditional AOD Limitation is placed on as in the given permitted largest beam speed of characteristic width by pattern displacement range.By calculating the limitation (the given shake width and AOD power limit of the function as AOD opereating specification) can determine under the given constraint most Good speed, thus optimize quantum of output.By the alternative of the conservative rate limitation of the setting worked in all cases, Ke Nengwu Method receives ground and reduces quantum of output.Allow to reduce the band of the galvanometer beam positioner using the third level filtering of AOD Beam Control Wide demand.The dynamic of the third level filter, and the function as range of deflection AOD optical efficiency variation, it is combinable with Limitation relevant to permissible process velocity when processing groove is set.In some instances, using automatic system and determination The processing of the optimum speed is to maximize process velocity.Shake is remained perpendicular to track by certain examples.Such example may Guarantee that the required linear extent is automatically kept, independent of process velocity.Which results in predictable and repeatable processing Quality, and permissible velocity arbitrarily changes on demand to optimize the quantum of output and power constraint in the beam positioner.Work as shake When light beam is to be formed in variable-width groove in workpiece material, for example, the dithered beam remains perpendicular to the tangent line of the groove And ignore the process velocity.Certain examples have used complete shake to go.Such example provides predictable processing result, Independent of the jitter points quantity for widening linear machining feature, and ignore the selected process velocity.The predictable processing The result is that it is useful, for example, when creating intersection between the features.By completing the shake row of integer, define well The end of the linear machining feature, allow with controlling depth becomes well in the intersecting of other features such as other grooves or pad Change.

Certain examples improve the processing efficiency of 2 dimensional region big on the workpiece.This can be by the scraping region come real It is existing, using shake to widen the processing stand.Which provides the fine resolution of the scraping area circumference, and provides and come from The abundant control of change in depth at the overlapping of the adjacent scraping access.Certain such examples using the overlapping controlled and High-resolution edge provides efficient scraping.Big scraping region is general in the application processed by LDA machine. In order to maximize quantum of output, optimizing the processing of these scraping features and keep the abundant resolution ratio of the edge definition in the scraping region It is useful.These examples make the use of the wider shake light velocity can be used in processing the scraping region, and keep Gao Pin simultaneously The edge of matter simultaneously shapes the overlapping between scraping access so that controlling depth changes well in the scraping region.

The attached drawing for being related to same components to wherein same tag number now makes label.For the sake of clarity, the label The figure number for the corresponding assembly that first numerical digit instruction of number wherein uses first.In the description which follows, in order to thoroughly understand Example disclosed herein and provide numerous details.However, be familiar with skilled worker it will be recognized that the example can be implemented and In none or multiple details, or there are other methods, in the case where component or material.Further, In some cases, the well known structure of not specifically illustrated or explanation, material or operation are in order to avoid obscuring each side of the invention Face.Moreover, the feature, structure or characteristic can be merged into any suitable mode in one or more examples.

Example may include each step, and can be embodied in will be by general service or special purpose computer (or other electronics Device) performed by machine-readable instructions in.Alternatively, these steps can be by including the specified logic for implementing these steps Hardware component is implemented, or by hardware, and the combination of software and/or firmware is implemented.

Example can also be provided as computer programming instruct, including have storage thereon can be used to computer (or other Electronic device) programming instruction non-transitory, machine readable medium, for implementing processing described herein.The machine readable medium can Including but not limited to, hard disk drive, soft dish, CD, CD-ROM, DVD-ROM, ROM, RAM, EPROM, EPROM, EEPROM, Magnetically or optically block, Solid State memory body device, or is suitable for storing other types media/computer-readable media of e-command.

I. it introduces

The present invention is described for the AOD control method in LDA system.The control extend the performance of the LDA structure with The quantum of output for being used for big scraping region is improved, and is used to support the high speed operation of groove processing.

The performance of the LDA system is to deflect the processing using acoustics-optical deflector (AOD) with flank speed (> 1MHz) Light beam is to provide radium-shine processing performance.Typically, which can deflect the light beam in the range of 5-20 spot diameter.The LDA system System can operate, for example, in vector, grating, in vector and scrape mode.

In vector pattern, the system is using groove processing as " line " in the workpiece.The AOD beam jitter is along one-dimensional (1D) line, in the axis for being orthogonal to the beam trajectory, artificially to widen the groove of the ablation.

In grating mode, region two dimension (2D) which deflects the machining beams to process in their processing spaces.This The space of a little grating regions and intensity pattern are generally arbitrary.The performance can be used, for example, being used to create between groove Intersect, or for creating the feature such as by pad.

In scrape mode, big region (being more than the AOD size) can be processed.It is real that shake can be used to widen line for this It applies.However, using specified processing to avoid the excessive pel at contoured edge, and be used to provide inclined fluence distribution with Reach good overlapping and intersection quality.

Following part describes these modes and lists the embodiment details of each example of the LDA system.

II.AOD processing Theory

In one example, the feature which manipulates the machining beams using AOD subsystem to create each dimension.It should Part describes the subsequent theory of the processing, guides system structure.

A. standardize for deep-controlled center fluence.

Material removal quantity is nominally proportional to fluence (for given radium-shine power grade), and thus controllable fluence With controlling feature depth.In addition, the phase interaction between characteristic size and fluence can be eliminated by the fluence of specified substitution dosage With.When prescribed dose, the remaining fluence in shake or grating region is based on the jitter points quantity and relative to the process points The characteristic size of size is the result of the Gauss point Energy distribution fallen in outside the shake net region.The result can be calculated And it compensates to eliminate the interaction.The system adjusts radium-shine power automatically dynamically to specify fluence to be maintained at the ditch this In the center of slot or grating region, independently of speed or shake (or grating) size of mesh opening.As a result, the elementary beam of the system Transmission setting (e.g., putting size, minimize shake (or grating) point and overlap) can be changed without influencing to process calibration result.

1. theoretical fluence (fluence)

Fig. 1 depicts the schematic diagram of the grid 100 of processing stand 110, can be by shake mobile beam (e.g., with vector Mode) or fixed grating pattern (e.g., with grating mode) Lai Shengcheng.The processing stand 110 each have respectively in Doa and Dca Coaxial (OA) and intersecting axle (CA) in XY point distribution.For the dithered beam according to this example, it is assumed that the shake figure (e.g., essentially permanently) case is repeated with the light beam in the movement (as by indicated by arrow 112) in the Y-axis.By Noa and Nca point in OA the and CA axis carrys out respectively coverage region A.In other words, Nca is in the intersection axis direction Point quantity.

I. grating fluence

It is assumed that the AOD residence time of each of radium-shine power P and Taod point, mean fluence (each unit area A ENERGY E) it is provided as

Fluence=E/A=P*Taod/ (Dca*Doa).

The arithmetic expression is useful in grating processing, is illustrated well by predictable grid interval.In this case, it can incite somebody to action " fluence scale " is defined so that fluence is converted success rate:

FluenceScale=(Dca*Doa)/Taod,

And the grating power (being indicated with watt) for being used subsequently to the specified fluence can be calculated as

Power=FluenceScale*Fluence.

Ii. vector fluence

In the case where the dithered beam, which is not fixed, because width is shaken in variable velocity and change. In this case, fluence calculating can illustrate according to dosage.It provides

Doa=V*Nca*Taod and

Width=Dca* (Nca-1),

Then

Fluence=P*Taod/ (Dca*V*Nca*Taod)

=(P/V)/(Width*Nca/ (Nca-1))

=Dosage/EffectiveWidth,

Wherein

Beam velocity (e.g., along the Y-axis in Fig. 1)=V,

Dosage=P/V, and

EffectiveWidth=Width*Nca/ (Nca-1)=Dca*Nca.

This is useful as a result, since it means that dosage and width can be used in (for ideal situation) (for shaking line) Or the fluence standardized in a given area is spaced with the known machined parameters of residence time (for grating region).Can will have The jitter feature of having time change point distribution and the spectra features of arbitrary point distribution merge to create by their each fluences etc. Intersection after grade standardization.

Dosage is controlled as long as noticing, there is no need to shake (the jitter points quantity Nd, between shake during vector is processed Every speed) details.This is useful, because the beam steering structure of the LDA system, which depends on, can arbitrarily change shake Interval and speed.

Iii. edge effect

Above-mentioned fluence model deviation theory, when the size of the dither pattern is smaller relative to the size.Such At small size, each point in power " leakage " grid 100 and reduce the fluence at the net region center.It should The result is that in the network both quantity of size of mesh opening and point function.For example, Fig. 2 is shown graphically according to an example as phase To in the EffectiveWidth (Weff) of the function of mesh width (being standardized by spot diameter Dspot) and the jitter points quantity Variation.It can predict and compensate the deviation in EffectiveWidth.

In Fig. 2, for the mesh width between 0 and about 1 that standardizes, which corresponds at 6 points, this is next A highest curve corresponds at 5 points, next one highest curve corresponds at 4 points, next one highest curve corresponds at 3 points, this is most Harmonic curve corresponds at 2 points.

2. fluence is standardized

The fluence at any radium-shine network of the equation calculation studied above be incorporated in the region processed respectively can Predictable result.Being somebody's turn to do " fluence standardization " has the groove and characteristic crossover for being used to form varying width.Two examples carry out it Description.

First example of fluence standardization is described in Fig. 3,4,5 and 6.Fig. 3 is described with figure according to an example Two cross features with different mesh-densities lattice.In Fig. 3, two characteristic crossovers, each has not Same shake grid interval.Pass through the round processing stand for indicating one of this feature and another feature is indicated by plus sige (+) Processing stand.For intersecting, one of these patterns may be, for example, the other part of groove and crossed grating pattern.Pay attention to the point Amplitude it is oblique to create the intersection slope so that tolerance is supplied to position error.Fig. 4 is described with figure to exist according to an example The modelling fluence of cross feature after amplitude standardization (not merging).It is each being scaled according to its dot density in Fig. 4 After the point energy of a grid, the fluence on each pattern is emulated.Fig. 5 describes the cross feature according to an example with figure Merging fluence.In Fig. 5, the fluence of two patterns is merged into the intersection to be used for relative smooth.

This mode is fine for the rectangular patterns effect of such rule.In more complicated crossing instances, such as groove With circular pad (common LDA intersects), which may include " protruding portion ", which is extended out and is reached The rectangle part that can be then handled as described above.For example, Fig. 6 is described according to an example with figure for having protruding portion The rasterizer grid (right side) and fluence of 612 circular pad 610 are distributed (left side).In the example in fig.6, which has big About 50 μm of length, can be used for cross over transition.

Second example is shown in FIG. 7, describes the note on varying width groove according to an example with figure The influence of gauge generalized.In Fig. 7, carry out tissue figure with three column, each column has the top figure of description dither pattern, description The intermediate pattern of corresponding fluence distribution, and description respective center fluence 710,712,714 (that is, when the groove width changes Fluence at the center of each groove) bottom figure.The column on the left side describe uncompensated, and the column of the centre describe width Degree compensation, the column on the right of this describe nonlinear compensation.In this example, it is scaled when simply by the shake mesh width When the dosage, whether non-linear in EffectiveWidth, which shows Fig. 2 and be still in when the shake width narrows, generates fluence Variation.On the contrary, Fig. 7, which is shown, reaches constant fluence using the non-linear fluence amendment when width changes.Such as institute in Fig. 7 Show, is based on width when the groove width changes using uncompensated center fluence 710 and using the center fluence 712 of linear compensation Degree variation.However, (e.g., being based on dosage and width parameter, wherein fluence=P* when correcting using above-mentioned non-linear fluence Taod/ (Dca*V*Nca*Taod)), which remains unchanged (that is, the curve 714 is maintained at the normative value “1”)。

After the fluence linearly to be corrected to the function for being used as and shaking (or grating) mesh width, constant dosage (power/ Speed) it then can be used to be kept separate from the fluence of speed.Thus vector-processing groove can become on width and speed Change and keeps center fluence constant.

Fluence linear compensation can be extended to two dimension to be used for grating region (e.g., for pad or other big features). However, the effect can be ignored, for being greater than the pad diameter more than 1.5* spot diameter.Since pad will usually meet this directly Diameter standard, in some instances, they may not be needed fluence compensation.

3. the warning for fluence modelling

Above-mentioned analysis is Utopian.It should refer to that possible influence is in the intersection that creation has controlling depth variation The several warnings for performance of uniting.

The material will have processing threshold value, thus fluence need not linear combining to predict the quantity of the ablator.This meaning Taste constant dosage processing may not generate it is constant as a result, because processing threshold function will reduce, when power with constant When speed at dosage increases.

Point distortion will make the fluence field distortion (including the effect deflected from AOD, especially at high deflection).

The possible untreated fluence on groove end is non-linear.Normally, groove end with another groove (in pad On stake, or in the suprachiasmal leg of grating) infall.As long as two intersection trench have identical on the intersection region Fluence slope, end influence nominally to cancel.

B. shake is extracted

Based on this as a result, the processing for extracting the shake and grating to be summarized to inferior grade details (e.g., jitter points of shake The scaling of quantity and they).This allows the system structure or component that can be easily modified, and (e.g., new AOD is designed, no Same optical layout) without the mode of influence user's (or programmer) specified jitter and grating processing.Simplify soft body structure, machine Device calibration, and application installation.

In the construction of the intersection, Ying little Xin avoids the variation in the fluence (thus depth) of the intersection.In example Example in, each intersection is customized based on the circumferential grooves size, each type is adopted as and intersects created customization Grating pattern (precalculates, e.g., use Matlab or other tools) offline.The processing can be fairshaped for producing Equipment.

For the LDA system, meet two targets by considering to form shake and the grating object of shaking dot grid, by Dot density and dimension attribute illustrate.By doing so, the fluence in shake and grating region on the workpiece can be calculated, and can Intersection grating pattern is generated based on the fluency calculated in this way.

1. shake figure

The details that shake geometry is not needed due to intersecting definition, the shake during vector is processed can be illustrated to extract To dot grid width and dosage.The placement of point, two in coaxial (along the velocity vector) and intersecting axle do not need to define Definition.On the contrary, the mesh width is converted into inferior grade jitter parameter by shake figure；Such as, the jitter points quantity Nd and the shake Width scale factor Kw.Notice that the concept of " shape " (Ks) is just being ignored now.

Processing calibration process described herein considers dither operation and fluence in the interface of the mesh width.The Nd and The details of Kw is encapsulated in the shake figure according to certain examples to avoid the complexity in the processing calibration steps.Pay attention to The shake figure can change, as fruit dot size or AOD range change.In some instances, identical shake figure is for adding Work calibrates and carries out being machined to ensure that consistent, the processing result after calibration during runtime.

The AOD shake for supporting the processing is preset during system calibration.The shake figure by jitter parameter with The interface at family, such as processing calibration, manuscript and sequencer.The shake figure establishes dither table using the following steps:

Firstly, determining that the AOD during system calibration deflects scale factor.

Then, effective spot diameter Deff in the system calibration data is set, is based on for 1:1 depth/width in length and breadth The minimum groove dimensions of ratio.The setting is only used for being arranged the guideline of the AOD table (in order to set conservative for the pitch Maximum value) and for fluence it is linear.The value of the Deff may need to be accurate in 10-20% in some instances.Deff's Lower value can be used for ensuring overlapping, but it will increase the jitter points quantity and is likely to reduced maximum speed.

Third, initialize 32 sequence dither tables (e.g., may be programmed in lock array (FPGA), as discussed below at the scene Equally).Each of these dither tables includes 1 to 32 points, each pitch having for 0.35*Deff.The dither table can Cover up to the jitter range of 10*Deff.This provides 10% deviation in the definition of Deff, and ensures overlapping pulse and create Uniform fluence distribution.

C. processing calibration and mapping

The example of the dose control is provided, which can be modified in the LDA system.In the LDA system Sample instance in, which includes inferior grade jitter parameter (Kw, Ks) and dosage (power/speed)-nominally every (although the actual dose unit may be arbitrary) in rice Joules (J/m) unit.The LDA system simplifies the processing sets up User interface, use is consistent, calibrated unit and the separation from inferior grade embodiment details.

The material processing in the LDA system: fluence (J/cm2) and shake or rasterizer grid can be defined by two parameters Size (for the width of vector sum shaving action, XY raster size) for grating.

1. processing calibration

Fluence standardization is for arriving the central area of shake (or grating) feature.The edge of this feature fluence not It is so easy to be typically canonicalized, because of its tail portion from the Gauss point.As a result, groove width is as the non-of shake mesh width Linear function variation, and be still calibrated in some instances.This is the function of the processing alignment features.Same process is used to school To the depths of features vs. fluence of given material.

Fig. 8 describes the exemplary process calibration test matrix 800 according to an example.The specified note of the calibration test matrix 800 Measure rate range and shake mesh width.Automatically processing one group of feature on this machine, (each is used in the parameter matrix Each entrance), and these features are measured based on external metrology workpiece.This feature width and depth results are imported should Processing calibration software, then sets up for automatically selecting machined parameters (fluence and grid for processing arbitrary dimension feature Width) " processing table " data bank.

2. manuscript and material table

Vector is processed, groove dimensions (width and depth) have determined the shake mesh width and fluence, passed through filling The manuscript of material table for the application.The manuscript implements the mapping based on the data provided from processing calibration table. Fig. 9 describes the examples material table 900 according to an example.Pay attention to the grey entrance in the material table 900 by the manuscript Input is not modified by the user.To processing competition subtle adjustment be it is obtainable with operation processing test or fine adjustments should Processing.

When loading this in application, using including all vector sum scraping features (groove width and depth) in this application The material table 900 is filled with the sizes of spectra features (pad diameter and depth).

If desired, the adjustment of fine adjustments can be inputted the material table 900 by the user.This may be in some instances It is depressing for product processing, but can be used for processing test and fine adjustments.If any processing calibration can not be obtained Table can be manually entered 900 entrance of material table.

One function of the manuscript be process calibration during the relative distribution information based on acquisition come interpolation this is required Material table 900.

III. system structure

The following part illustrates system structure and implements certain examples of the component of AOD operation.

A. data process and processing

Figure 10 is to describe to be used to that the processing of the simplification of the AOD and data process structure is ordered (to be also referred to as herein according to an example As " AOD processes structure " 1000) block diagram.It includes that system controls computer (SCC) 1010 that the AOD, which processes structure 1000, is added Work cluster 1012, AOD front end-plate (AFEB) 1014, AOD driver 1016,1018 and AOD1020,1022.It specifically discusses as follows Equally, the radium-shine processed and applied of the SCC1010 preprocessing is to construct individually processing segmentation and related processing parameters.It should The segmentation data is sent the interface that (e.g., with 200kHz data speed) arrives the processing cluster 1012 by interface 1023 by SCC1010 1024.The processing cluster 1012 includes digital signal processor (DSP) 1026 and field-programmable lock array (FPGA) 1028.Technology Personnel will from there disclosure understand can also be used other kinds of processing logic (e.g., compensation or be not the DSP 1026 and/or FPGA 1028).

The DSP 1026 calculates specific light velocity track and jitter parameter (e.g., with 1MHz data speed).Then by the money Material is transmitted to the FPGA 1028.As discussed in detail below, which calculates high speed AOD order and (e.g., reaches 4MHz data speed), the AFEB 1014 can be sent to by SSP1 serial link.The AFEB 1014 is by the AOD order It is converted into being sent the parallel data word with the AOD driver 1016,1018.AOD driver 1016,1018 subsequently generates radio frequency (RF) driving signal operates the AOD 1020,1022 to control optical light beam deflection (by RF signal frequency) and amplitude (passing through RF signal amplitude) passes through the AOD unit.The exemplary operations of the AOD 1020,1022 and other optical modules are referring to figure 10A, 10B and 10C are discussed below.

Figure 10 A be according to an example include AOD subsystem 1042 and galvanometer subsystem for shaking radium-shine light beam The block diagram of 1044 system 1040.Technical staff disclosure will recognize from there, and other kinds of locator can be used System.For example, FSM can be used in positioning subsystem.The system 1040 includes radium-shine source 1046, for providing machining beams 1048 To the AOD subsystem 1042.In one example, which includes that pulsed laser source makes the machining beams 1048 to include A series of laser pulses.In another example, which includes that the radium-shine source continuous wave (CW) makes the machining beams 1048 include CW radium-shine light beam.In certain such examples, the AOD subsystem 1042 is by deflection between discrete (" pulse ") Every place's machining beams 1048 to generate laser pulse from the CW radium-shine light beam.

As discussed above, the AOD subsystem 1042 is by the machining beams 1048 at the AOD deflection angle 1050 The first rank light beam 1049 and the zero-order beams 1051 of the machining beams 1048 deflect into optical beam dump 1052.The system 1040 It can further comprise fixed mirror 1054 so that the first rank light beam 1049 is deflected into the galvanometer subsystem 1044 and scanning lens 1056 to focus on laser light beam spot 1058 on workpiece 1060 or interior.Here referred to as can be by the output of the scanning lens 1056 The radium-shine light beam 1061 of focusing.

In one example, which may include for providing in (e.g., jitter direction) in a first direction back and forth The single AOD of deflection, and the galvanometer subsystem 1044 is provided along the deflection in the second direction of machining locus 1062.Scheming In example in 10A, every a line (being shown along the X-axis) of AOD jitter points is perpendicular to the machining locus 1062.In order to provide Speed and multipurpose, however, the AOD subsystem 1042 in the example described in Figure 10 A is provided relative to the workpiece 1060 Surface is deflected along the 2-D of X-axis and Y-axis.In this example, which can be referred to as being parallel to the machining locus 1062, it can The X-axis is referred to as perpendicular to the machining locus 1062.Thus, which can be referred to as to the jitter direction.The machining locus 1062 can correspond to a direction, for example, the system 1040 scribing line or cutting groove 1064 are (e.g., in the galvanometer subsystem 1044 Control under) to the direction in 1060 surface of workpiece.

In order to provide described 2-D deflection, which includes the first AOD 1020, be used to by this first Rank light beam 1049 deflects into first direction and the 2nd AOD 1022, for the first rank light beam 1049 is deflected into second party Xiang Zhong, when galvanometer subsystem 1044 moves the beam axis along machining locus 1062.It in other words, will be by AOD The mobile superposition of light beam spot position provided by font 1042 is in the light beam spot position as provided by the galvanometer subsystem 1044 Movement on.As shown in Figure 10 A, which may also comprise the first galvanometer mirror 1066 and the second galvanometer mirror 1067 deflecting the first rank light beam 1049 in the X-axis and Y direction relative to 1060 surface of workpiece simultaneously.

The orientation of AOD deflection may be misaligned to the yawing axis of the galvanometer subsystem 1044.It generally, can be to this AOD deflection command is using coordinates translation obtained AOD deflection is aligned in required coordinate framework.The coordinates translation may also It is the function of speed, rotates AOD deflection coordinate framework remaining perpendicular to AOD light beam deflection by the galvanometer subsystem Machining locus defined in 1044.

Using including the AOD subsystem 1042 in the system 1040, several operation modes are enabled.In one example, it grasps Operation mode includes the ability for shaking the machining beams 1048 effectively to widen the laser light beam spot 1058 at the workpiece 1060.It changes Sentence is talked about, and shaking the machining beams 1048 includes spatially positioning a series of focusing laser light beam spots 1068 to create and have greatly In the geometrical characteristic of the size of each laser light beam spot 1058 focused by the scanning lens 1056.In order to be described, scheme 10A shows being somebody's turn to do from the viewing of the surface of the workpiece 1060 when processing the groove 1064 in 1062 direction of the machining locus Shake laser light beam spot 1068.Thus, for example, having to give the series shake laser light beam spot 1068 of repetitive rate with lower arteries and veins Rush a series of result for larger diameter laser light beam spots that repetitive rate is continually used in 1062 direction of machining locus.

In some instances, the AOD1020, the 1022 each sound fields that can update them (are inserted using new sound wave The optical aperture), with about 0.1 μ s to the sequence of about 10 μ s.It is assumed that the example renewal rate of about 1 μ s, can quickly update The position of the machining beams makes several shake laser light beam spots 1068 overlap during processing.The shake laser light beam spot 1068 The dimension (e.g., along the X-axis or jitter direction) of the vertical machining locus 1062 can be overlapped in widen the feature just processed (e.g., the groove 1064).As shown in Figure 10 A, which can also overlap in the side of the machining locus 1062 Xiang Zhong.It is that the machining locus 1062 can be with the machining locus according to certain examples to keep the dithered beam normal orientation Shaker shaft is consistently adjusted when 1062 angle change.It is endowed with compensating in the jitter points in addition, can adjust the shaker shaft Function of the angle as the machining locus speed on line.

Other than the light-beam position shake relative to 1060 surface of workpiece, or in other instances, the AOD subsystem System 1042 can be used to change the light distribution in the shaker shaft.Along the light distribution of the machining beams 1048 of the shaker shaft Manipulation make it possible to the forming of the cross section of the processing groove 1064.For example, rectangle, U or V-arrangement cross section can be used to process The groove 1064.In the case where formed features such as skew back panel can have for such as cross conformation.The forming resolution ratio is potentially based on The basic point size, and the light distribution of the forming may be the light distribution of dither pattern (position and the intensity) and the point The convolution of (such as Gauss or other distribution shape).Feature is formed, it can be for example, by overlapping along the shaker shaft at certain Selected amount of the pulse (e.g., can be used in two or more pulses same is position) to remove target material at a little positions, and/ Or function of the power magnitude by modulating the laser pulse as the inflection point along the shaker shaft.

Other than the feature shaping along the shaker shaft, or in additional examples, which be can be used to The power, which is controlled, as the function of the position along the machining locus 1062 allows the phase of " terminal " of machined linear character With forming.Controlling the power can also be used for application such as cross conformation as the function of the position along the machining locus 1062 In.Using for the AOD subsystem 1042 enables algorithm for power modulation to occur to make in very high speed (e.g., in microsecond rank) The subtle control (e.g., having the characteristic size between about 5 μm and about 50 μm) of the light distribution adds in height It can be possibly realized at work speed (e.g., in the range between about 1m/s and about 5m/s).

Other than the deflection of Gaussian beam, certain examples also deflectable light shaped by traditional beam-shaping technology Beam includes for example, diffraction optical assembly (DOE).For example, Figure 10 B is the box according to the Beam shaping system 1070 of an example Figure.The system 1070 includes the AOD subsystem 1042 (using the first AOD 1020 and the 2nd AOD 1022), the zeroth order light Mirror 1054 in beam collector 1052 and Figure 10 A.The system 1070 further comprises that diffraction optical assembly (DOE) 1072 is used for Beam-shaping and optical module 1074 (e.g., imaging optics, galvanometer mirror and scanning lens).In order to scheme for describing The first rank light beam 1049 in 10B is shown as on 1050 range of AOD deflection angle.It, will be by this in the example described in Figure 10 B The first rank light velocity 1049 that AOD subsystem 1042 is deflected is relayed to the DOE 1072 (by the light beam by relay lens 1076 Fulcrum is imaged on the DOE 1072) to keep the first rank light beam 1049 to ignore at the center in the aperture of the DOE by the AOD The AOD deflection angle 1050 that subsystem 1042 is assigned.The subsequent DOE 1072 can by assign before additional wave phase distortion come (the same as is common for such beam-shaping DOE) is shaped to the beam intensity.Which is advantageously possible for can be used for example In the case where square light distribution is distributed with to form more unified shake fluence deflecting and abut larger, the forming light velocity.The party Formula may also be conducive to be used to be formed required feature (for example, micropore in dielectric material) enough in smallest number laser pulse In situation.In such a case, the grating application of Gaussian pulse may be inefficient for the light distribution for using forming, Even high speed AOD deflection can be required for the High-speed Control of intensity processing stand position of the forming.

In additional examples, identical relay lens configuration can be used to cut out the AOD deflect light at the scanning lens The deflection of beam.This may be for needed at least two reasons.Firstly, it may be desirable to the fulcrum of the light beam is relayed to the inspection stream Meter scanning mirror (eliminating the deflection of light beam side) keeps the light beam in the clear aperature of the galvanometer mirror and scanning lens with (a) The heart is cut out to avoid light beam, and (b) avoids for the light beam being displaced from the scanning lens entrance pupil center, because of such position There may be the diagonal beams at workpiece surface for shifting.Second, it may be desirable to, in order to generate light intentionally at the work top Speed tilts and assigns at the scanning lens light beam deflection of side.Diagonal beam can have in the application of certain Gauss laser drillings Conducive to the side wall that the creation in machining feature (such as micropore) is precipitous.

Figure 10 C is the block diagram according to the system 1080 of the light beam 1082 of the offer inclination processing of an example.The system 1080 include the AOD subsystem 1042 (using the first AOD 1020 and the 2nd AOD1022), the zero-order beam collector 1052, and the mirror 1054 in Figure 10 A.The system 1080 further comprise relay lens 1076 and optical module 1074 (e.g., at As optical component, galvanometer mirror and scanning lens).In order to be described, the first rank light beam 1049 display in Figure 10 C is existed On the range of AOD deflection angle 1050.As shown in figure 10 c, saturating from the scanning with spacing 1084 by correctly designing The relay lens 1076 (e.g., the scanning lens 1056 in Figure 10 A) of Jing Chu has that the AOD subsystem 1042 is deflected Single order light beam 1049 also can be deflected laterally to create the diagonal beam 1082 at 1060 surface of workpiece.It can be controlled for The light beam tilt quantity of the given deflection of the processing stand uses the AOD 1020,1022 with basic by (a) at the workpiece 1060 Side point deflection of the upper creation at the workpiece 1060, and the relaying of change to the scanning lens (e.g., scanning lens 1056) is saturating 1076 optical component of mirror and spacing 1084, or (b) coordinate the galvanometer (e.g., the galvanometer 1066,1067 in Figure 10 A) and AOD 1020,1022 make any side light beam at the scanning lens deflect (and the thus arbitrary beam at the workpiece 1060 Inclination) it can independently be assigned from the required side point deflection at the workpiece 1060.

The further details of forming technique are disclosed in the part for being designated as " example A OD control example " below.

When beam jitter is highly effective and flexible for fluence distribution needed for generating, reach the replacement of shake (but sometimes It is tightened up) mode includes that at least one changes the radium-shine light beam by the way that linear frequency modulated waveform is used for the AOD 1020,1022 The focusing of point 1058.Using linear frequency modulated waveform, the real-time frequency of the sound wave is in the optical manufacturing light beam for passing through the AOD crystal Linearly change in 1048.The linear change of the real-time frequency of the sound wave, which has, focuses item applied to the processing for uniaxial (astigmatism) The effect of light beam 1048, rather than mobile radium-shine light beam 1058 in discrete step.By the way that linear frequency modulated waveform is all used for AOD1020,1022, can be symmetrically by 1058 defocus of laser light beam spot according to certain examples, thus increase in the workpiece 1060 The size of the point at place.Which may be for, for example, the pulse recurrence frequency may be not high enough in can provide wherein Good pulse is overlapped at the workpiece 1060 and repeats to mould to avoid the lower of the Strength Changes when widening the groove 1064 In the case where glue is radium-shine.Linear frequency modulation can also be used for defocus laser light beam spot during using the procedure of processing of lower fluence.Example Such as, radium-shine processing (e.g., be formed between semiconductor core on piece or interior integrated circuit cross) may include cutting through overlapping metal First procedure of processing of (e.g., copper) layer processes the second procedure of processing of following dielectric layer followed by reduced fluence is used.With It uses the access of two radium-shine light beams, and not as good as an example usability frequency modulation, with defocus, the laser light beam spot individually leading to Two floor are all processed in road.

B. wobble structure

Basic jitter 1. (or grating) table structure

Figure 11 is the block diagram for indicating the AOD control data process according to the implementation of an example in FPGA 1028.It should It is noted that often convertibly using shake and grating.They are all similarly performed；However, shake is related to one-dimensional light beam Deflection and grating is related to two-dimensional deflection.The FPGA 1028 includes one or more dither tables 1110, shapes table 1112, linearisation Table 1114,1115 and waiting time adjustment 1116.

Two groups of jitter points are loaded into dither table 1110.The dither table 1110 operation is loop buffer, can be ad infinitum It is addressed.Many dither tables 1110 are stored in the FPGA 1028.Each dither table 1110 is identified by address and length. The loop buffer of 1028 adjust automatically of FPGA addressing is to accommodate the length for specifying table.

During grating or dither operation, which is read and is determined by transition matrix to scale and rotate Shake (or grating) pattern forms the frequency order for arriving two channels ADO.

A pair of of additional frequency deviation is added on to the shake order of the conversion, is provided to the shake order using relevant mark The command vector of title.

The set of the linearisation table 1114,1115 generates AOD amplitude order based on the AOD frequency order.

It is similarly Figure 11 and provides following definition.

F0:AOD frequency order, axis 0.

F1:AOD frequency order, axis 1.

Fnom: the nominal AOD frequency order for zero deflection.

Fd [1..Nd]: the deflection frequency collection including above-mentioned " dither table " 1110.

Nd: the quantity (e.g., the jitter points quantity) of deflection frequency point.

Kw: shake width scale factor.For non-jitter Kw=0 (nominal machining beams).

Kp: power command scale factor.

Ks: intensity shapes factor.

Atten: decaying order.

The interface 2.FPGA

Data is transmitted to the FPGA 1028 from the DSP 1026 by (e.g., 1 μ sec) when each Tcmd updates.

3.DSP/FPGA is synchronous

During processing, the DSP 1026 continuously by data process to the FPGA 1028, and the FPGA1028 by they Data transmission is synchronous with AOD control execution.Following sequence can be used to complete in this, it is assumed that example renewal time (Tcmd) is 1 μ sec.Figure 12 describes the timing.

Figure 12 depicts the timing diagram of the example sync of the DSP and FPGA according to an example.As shown in Figure 12, should 1210 FPGA 1028 are written in ten groups of control datas (being sufficiently used for two 5 μ sec DSP periods) by DSP 1026.The DSP " synchronization " register of 1026 settings 1212 in the FPGA 1028 is to notify its data to prepare for processing.The FPGA 1028 start information processing 1214 at next 5 μ sec DSP interrupt 1216, as as indicated by the dotted line 1218. (FPGA 1028 and the DSP 1026 receive identical interruption).The DSP 1026 is by new data during each DSP period It loads at 1220 to the FPGA 1028.The FPGA 1028 handles the data based on first in, first out (FIFO).In each DSP Galvanometer order data is calculated during period, but is not used in the galvanometer controller at next DSP period.The galvanometer 1066,1067 and AOD1020,1022 thus share the identical timing reference.The DSP 1026 is relative to the galvanometer order Data postpones the AOD order data to consider the various signal processings delay between the galvanometer and AOD control.The delay The beginning for being incorporated in baseline file transmission (being shown at 1210) and the data carried out by the FPGA 1028 execution (is shown in At 1214) between two cycle delays adjustment.

4. come down parameter

A unshowned parameter is " landslide " parameter (Kb) in Figure 11, and it is fixed can be included in the LDA thrashing In justice.The parameter is used to change power used on the dithered beam width to consider in the circular arc (the similar match at landslide Road bend) speed difference between inside and outside edge.This is used for the dither amplitude as trembling by that will add scale factor Dynamic position function is realized, " forming " parameter being used in some instances is similar to.

Figure 13 depicts the schematic diagram of the processing situation using landslide.Figure 13 describes wide circular arc 1310.This is wide Circular arc 1310 includes relatively average radius R, may be bright with internal arc length Ri to the ratio of external arc length Ro Aobvious ground is different from 1, and the fluence on groove width W is caused to change.

The ratio of mean radius R to characteristic width W is defined as landslide ratio

Rb=R/W.

Then the ratio of the outwardly and inwardly speed is

Vo/Vi=(2*Rb+1)/(2*Rb -1).

In order to adjust the radium-shine power on the shake width, the conversion of the power scaling is defined as by " landslide parameter " Kb The function of dither positions.Figure 14 is described with figure according to the example of example landslide parameter scaling.Landslide parameter Kb -1 to+ It converts, and can be defined as between 1

Kb=(Kstart-Kend)/2,

Wherein

Landslide amplitude scaling of the Kstart=at shake period starting, and

Landslide amplitude scaling of the Kend=at shake end cycle.

The landslide parameter is relevant to the landslide ratio and passes through

Kb=1/ (2*Rb).

For Kb=0, the scale factor in the jitter range is 1 (no to influence).The extreme case (| Kb |=1) Place, is reduced into zero for an endpoint of the jitter range and amplifies at another port by 2x.When processing traditional circular arc, Kb can be set to median (e.g., to be less than about 0.5).

Landslide scaling at the center (that is, center line of the light beam of the shake) of the jitter range is unaffected.Thus The center linear heat generation rate (Pnom) holding do not change, and by the inner edge power attenuation (compensate the lower beam velocity) and by this it is outer Edge power (Pouter) amplifies (compensate the high light Shu Sudu):

Pouter=Pnom* (1+Kb).

The LDA design rule is pointed out in some instances, which should be restricted to be greater than about 2.0 to keep closing Manage velocity rate and power scaling.During processing, the power of the amplification is checked to verify it without departing from function obtained by maximum Rate.Speed in the circular arc is set to ensure the power of the amplification without departing from power obtained by the maximum.Due to the third level The performance of section, which, which is reduced, can occur immediately in the end that the circular arc is segmented.

5. example dither operation

One simplified example is described in Figure 15, and the shake timing in 16 and 17 and order update.Figure 15 is described with figure The example XY light-beam position of dither operation according to the example.Figure 16 describes showing according to the example dither operation in Figure 15 with figure The example X and Y light-beam position vs. time.Figure 17 describes example light-beam position and the processing of the example dither operation in Figure 15 and 16 The table of parameter.In this example, Taod=Tcmd=1 μ sec.Using a dither table (Nd=5, Td=5 μ sec) by initial ditch Slot cutting.Change dosage (Kd) at the beginning of the cutting.The landslide parameter (Kb) is modified during the circular arc to be somebody's turn to do to standardize Fluence on curve inner edge and outer rim.

Groove in this example starts increased transient sublevel 1510 including the wherein groove width.In this example, The transient sublevel 1510 includes the 13rd and the 14th shake row.However, in other instances, the shake of different number can be used Row (e.g., four, five, six, or more shake row can be used in the transitional region in).Most start and (e.g., starts from the 13rd row), it will Larger dither table (Nd=7) compression (Kw < 1) is to match the previous segmentation.Kw then increases on two shake rows, until Kw =1 (entire width).In other words, as shown in Figure 17, the Kw=0.76 in the 13rd shake row, the tenth Kw=0.87 in four shake rows, the Kw=1 in the 15th shake row.

Dosage (Kd) is also modified in the dscreasing width parameter.(e.g., which opens when compressing the shake interval End originates in the 13rd shake row), dosage is reduced to consider that the biggish pulse overlaps.In other words, for the mistake of Nd=7 Cross the outside of segmentation 1510, Kd=1.4.(wherein the dither table changes from Nd=5 to Nd=for the first time within the transient sublevel 7), however, Kd is reduced to 1.06 in the 13rd shake row.Then, Kd is increased in the 14th shake row, again It is secondary to increase to before the 1.4 of the 15th shake row.

Shape remains unchanged in this example, although general its may be changed in the mode similar to Kd, Kw and Kb.It is right In LDA system in one example, ignore shape.

Notice that machined parameters (Kw, Kd, Kb and Ks) and dither table selection (Nd) are shaken row for each and kept not Become.As indicated earlier, which implements the parameter and updates timing；The DSP 1026 is capable of providing all processing The regular update of parameter simultaneously allows the FPGA 1028 to use them when suitable.

C.AOD coordinate framework and calibration

The deflection of AOD light beam is with the deflection co-ordination of galvanometer light beam to generate final workpiece surface light-beam position.The AOD The calibration of deflection, and the interaction of its deflection with the galvanometer light beam during third level positioning, shake and grating are useful In keeping local repeatability (e.g., for support cross conformation) and control is shaken and grating feature dimension.

In one example, the AOD subsystem 1042 and the galvanometer mirror 1066,1067 each deflect the machining beams angle Degree, before it enters the scanning lens 1056.The deflection of any AOD beam angle is added on the inclined of the galvanometer 1066,1067 Turn, and thus the AOD deflection command can be considered being equal to " original " order-of galvanometer that it is " beam angle " coordinate.It can incite somebody to action AOD deflection takes as to be added to " beam angle " order of " beam angle " of the galvanometer 1066,1067 order.In certain examples In, only in heap AOD light beam deflection coordinate framework scaling and rotation to match the galvanometer coordinate framework (as described in Figure 38) Later, this may be correctly, because optical series layout may create the rotation between two axis, and the AOD Deflector 1020,1022 has unique scale factor.

In the presence of the second-order knot of the displacement for the virtual AOD deflection fulcrum being formed at the galvanometer mirror 1066,1067 Fruit.Such as AOD 1020,1022 deflection light beam is the same, side light-beam position slight shift in the lens entrance pupil, Lead to lesser additional distortion.The error, is contemplated to be sub-micron, can be ignored in some instances.

There is the AOD 1020, the concept for " increasing angle " used in 1022 has for considering third level section, shake When with raster manipulation, there is the meaning about the required calibration and need to meet the runing time amendment of system performance goals.

1.AOD conversion

In some instances, the seat by specifying the type AOD processing is originated to the command signal of the AOD subsystem 1042 Mark conversion.The following discussion refers to several coordinate systems." ideal " coordinate is the coordinate of the workpiece surface of calibration.It can pass through The eyeglass mesh standard in the XY stage is to these definition.The term " required " " nominally " can be used for replacing " ideal herein "." original galvanometer " coordinate is used to order the galvanometer servo loop.It is logical that " original AOD " coordinate can be used to the order AOD Road.

Following conversion is incorporated in this: TMframe, TMfield, TMdither and TMaod.TMframe flows original inspection Coordinates translation is counted to original AOD coordinate.By the way that TMframe to be used for the original galvanometer coordinate of one group of growth (from nominal Galvanometer position) and the AOD order of formation creates and is optically equivalent to increase what galvanometer deflected about the nominal position The deflection of AOD light beam.Thus the AOD becomes " virtual galvanometer " after TMframe to be used for original galvanometer order.This turn It is fixed for changing for given optical layout, and is not changed as eyeglass grid is calibrated.By the AOD position correction program come Calculate the conversion.

TMfield is local scanning field distortion conversion.It (sweeps ideal (or required) coordinate of one group of growth from this Retouch some nominal positions in field) it is converted into the original galvanometer coordinate increased." forward " (preferably arrive original galvanometer) and " opposite " (original galvanometer is to preferably) is all used.The conversion is the function of scanning field position.It is with eyeglass grid school It is quasi- and change, because the calibration defines ideal coordinate framework.It can be calibrated from galvanometer and calculate TMfield from data.

TMdither is one group of transition term, is transmitted to the DSP 1026 for processing the phase in vector from the SCC 1010 Between TMaod calculate.TMdither is each vector or shaving action segmentation calculating by the SCC 1010, and is scanning field position Set the function with beam velocity.

TMaod be the ideal AOD data being stored in shake/grating table inside the FPGA 1028 is carried out rotation and The conversion of scaling.In vector or shaving action, the ideal shake vector perpendicular to the ideal trajectory vector is maintained.In light In grid processing, two-dimensional grating number is scaled and is aligned according to the needs for intersecting processing or general grating pattern information.In vector Or during shaving action, by the DSP instant computing conversion, turned based on the beam trajectory velocity vector and the TMdither It changes.TMaod conversion is once calculated by the SCC 1010 during grating processing.

2. scanning lens field distortion

The focus is about the seat at the field distortion for being formed in the scanning lens 1056 and galvanometer mirror beam delivery system Mark framework and Calibration equation.In ideal scanning lens, by the angular deflection of the upcoming light beam to generate in the work Point displacement at part surface.For the telecentric scan lens (" F- θ " lens) of good design, if the light beam is with the lens entrance The center of pupil is core, and point displacement is proportional to angular, and does not need any calibration.However, at the entrance pupil The physical constraint of encapsulation of galvanometer mirror 1066,1067 prevent the light beam practical using the ideal point as core；It is deflected in light beam Light beam conversion has occurred in period.This create scanning field distortion patterns, as shown in Figure 18.

Figure 18 describes the scanning field distortion pattern in sample instance with figure, and wherein the LDA system includes F/18 100mm scans field lens.In typical radium-shine system of processing, maps the distortion pattern and make that correction term can be applied to the light Beam positioners, (nominally) the undistorted pattern being formed at the workpiece surface.The output of such calibration conversion is at this Order-in " original " coordinate framework of beam positioner is equal to the mirror angle in 1066,1067 situation of the galvanometer Order.In the LDA galvanometer controller subsystem, nominal scale factor (e.g., the 2* focal length of lens) is used, work is formed in Original directive in the unit of part surface μm.In the example using FSM positioning system, positioned at the entrance pupil of the scanning field lens The FSM at place will have the relatively fewer distortion that will be corrected, compared with galvanometric distortion.

Note that practical field distortion measured in system is several in the optical series (galvanometer block+scanning lens) The combination of what distortion and galvanometer positioners error (deviation, scale factor are non-linear).The optical field distorterence term for inspection stream and AOD deflection is all common, and identical correction term can be used.However, any school in the galvanometer 1066,1067 itself Quasi- error (linearity error and scale factor (SF) shake based on angle) is also included in scanning field calibration item, and is being used for AOD creates calibration error when deflecting.However, the galvanometer errors are smaller, the scale factor (SF) with < 0.1% is non-linear Error and temperature jitter, may be formed at the third level AOD deflection for 100 μm < 0.1 μm of galvanometer and AOD deflect it Between mismatch.However, the nominal scale factor tolerance on galvanometer 1066,1067 is possible larger (e.g., 20%), The calibration process is influenced, as described below.

The scanning field distortion pattern produces the local distortion as the variation of scanning field position function.For example, Figure 19 schemes Describe (err) vs.X and Y position X and Y scan field distortion error according to a sample instance.Figure 20 describes basis with figure The X and Y scan field distortion scale factor (SF) of sample instance in Figure 19 and vs.X and Y positions of rotation error.Figure 21 figure It describes and is distorted according to the example local position geometric corrections (PGC) of the spectra features of the example in Figure 19 and 20.It can be by conduct The X of XY positions and the slope of Y error are treated as local scale factor and rotation error, e.g., PGC distorterence term.If this A little errors are sufficiently large, and PGC distortion in the local can create the unacceptable error in grating pattern.For example, it is contemplated that be located at (- 50mm, -50mm) field position 200x200 μm of grating pattern.Local PGC is distorted (picking up from Figure 20) and will generate shown in Figure 21 Distortion pattern.The X error reaches 5 μm due to the biggish rotation error in the pattern corners, may be sufficiently large to produce The raw unacceptable change in depth in cross conformation.

The scanning field distortion pattern is to influence the LDA beam positioner at least in a manner of three kinds: in all moulds

During formula, when third level section defines AOD and galvanometer order to generate nominal beam trajectory；In vector Or during scrape mode, when scaling and rotating AOD shake order；With during grating mode, it is distorted when potentially possible When larger grating region.

3. the AOD calibration correction during third level section

Figure 22 describes the third level section subsystem 2200 according to an example with figure.In third level section, in the inspection Divide the positioning of light beam between flowmeter subsystem 1044 and the AOD subsystem 1042.Third level section is related to the AOD subsystem 1042 are used as third level locator (e.g., to the supplement in XY stage and the galvanometer subsystem 1044).In United States Patent (USP) 6,706, Example radium-shine light beam third level locator is described in 999, is assigned to agent of the invention, and it integrally leads to accordingly Reference is crossed to be incorporated herein herein.As disclosed herein, allowed using the third level section of the AOD subsystem 1042 with high speed (e.g., use with about 1 more newly arriving for μ s improve timing resolution) describes the profile of the beam path, wherein based on it is discrete when Sequence boundary proposes AOD order.The third level section subsystem 2200 includes profiling filter 2204, Delay Element 2206 and is subtracted Musical instruments used in a Buddhist or Taoist mass 2208.

Figure 22, which describes to correspond to, to be needed to be cut into the example beam distribution 2210 of groove of workpiece (it can also be here It is called example " light beam order ").The example beam distribution 2210 includes can with high speed using the galvanometer subsystem 1044 It can be difficult to the zig zag tracked.After through scanning field calibration conversion 2203, which is improved to the Three-level filtering device 2205 comprising the profiling filter 2204 and the Delay Element 2206.The profiling filter 2204 includes filter Except the low-pass filter that may be difficult to the high-frequency content tracked for the galvanometer subsystem 1044.The profiling filter 2204 output can be used as galvanometer order (galvanometer control signal), as shown in position distribution 2212.Figure 22 describes the position The amplifier section 2213 of distribution 2212 is set, it illustrates relative to the physical location as provided by the galvanometer subsystem 1044 The position 2216 of 2218 order.The AOD subsystem 1042 is used to correct in the position of the order 2216 and the physical location Difference between 2218.

In one example, which includes that (IIR) filter is responded in unlimited pulse.In another example, The profiling filter 2204 includes that finite impulse responds (FIR) filter.FIR filter itself, which has, is used for optional frequency range The fixed delay of interior signal.However, technical staff disclosure should recognize from there, it is possible to use other kinds of filtering Device.The Delay Element 2206 postpones the example beam distribution 2210, passes through the substantially phase introduced by the profiling filter 2204 Same retardation.The subtracter 2208 by the output of the profiling filter 2204 from output place of the Delay Element 2206 subtract with Obtain the high-frequency content removed from galvanometer order place.The high-frequency content of the subtracter 2208 can then be exported and be used as AOD command signal is used to control the AOD subsystem 1042.Figure 22 describes example A OD position command distribution 2214.Although not showing Out, differential can be used in position command distribution 2214 to calculate corresponding speed and acceleration command distribution.

The example light beam order 2210 is the required track of the light beam on the workpiece surface, in the application of planar alignment conversion Later, using " required " coordinate.As discussed above, by (the light-beam position as order of example beam distribution 2210 Signal) it is supplied to scanning field calibration conversion 2203.The data is filtered so that the track is divided into low frequency and high fdrequency component, is allowed The AOD subsystem 1042 tracks high frequency, low amplitude value order, and bandwidth limitation, amplitude order are transmitted to the galvanometer subsystem System 1044.Using scanning field calibration conversion 2203 with generation " original galvanometer " coordinate.Since this generation is filtered in the third level Before the 2205 pairs of orders segmentation of wave device, the output of the third level filter 2205 is galvanometer and AOD component, each of these In same original galvanometer coordinate.

If calibrate the AOD subsystem 1042 with deflect use the original galvanometer coordinate framework light beam, do not need for Any further scanning field calibration conversion is taken in AOD third level displacement.This be it is useful, do not need this since it implies Ground AOD field distortion amendment.In other words, the scanning field distortion is already have accounted for when using scanning field calibration conversion 2203 to imitate Fruit.

Another explanation of which is that the galvanometer order is moved away from the required order by the third level section filter, In original galvanometer.The AOD subsystem 1042 simply provides compensation displacement to form the galvanometer light beam angular displacement.

Then conversion (scaling and rotation) will be exported in the AOD order in " original galvanometer " coordinate should be " former to generate Beginning AOD " deflection command.The conversion is referred to as should " TMtert " conversion.

TMtert conversion is kept different (shown in Fig. 1) from the TMaod conversion for being used to modify shake, at least two Reason.First, the scaling inside TMtert cannot be used during shake, because it all corrects AOD and galvanometer SF, and because Rather than it is processed independently of the shake of galvanometer movement.Rotation item in second, the TMtert is fixed and independently of this Velocity vector angle, when for the TMaod shake conversion (using inside the FPGA 1028) changed with the trajectory angle.Thus will TMtert conversion is used for the track AOD data, before being transmitted to the FPGA 1028, and the data not by TMaod into one Step influences.

It is used for " original galvanometer " coordinate and also provides TMtert conversion that AOD error correction item is increased to the third The chance of grade AOD data.This is convenient, because (it is filtered to generate AOD error correction money the galvanometer controller error Material) it is in original galvanometer coordinate.

The calibration data process of the formation is outlined in Figure 51.

The processing discussed above for being used only for being supplied to the beam center track position of the third level algorithm filter.With tremble Dynamic calibration effect relevant with grating is slightly different, as described below.

4. the AOD calibration correction during grating pattern structure

As described above, scanning field PGC distortion in the local may be large enough to need the grating pattern as field position function Local calibration amendment.Notice that the situation is different from above-mentioned third level section, because AOD deflection is not carried out to replace in advance The galvanometer of compensation deflects (it means that pre-compensation of the AOD deflection command)；On the contrary, only field distortion compensates for the light The central point of gate pattern.The AOD deflection for processing the grating pattern is not compensated.

PGC amendment in the local is converted from the scanning field calibration, and can be by from the SCC during or before grating processing 1010 are transmitted to the DSP 1026.PGC conversion in the local is merged with other AOD conversion and is used to shake angle and AOD coordinate structure Frame rotation, as described under " vector processing is summarized " later.

5. the AOD calibration correction during shake

In the case where shake, the AOD local PGC distortion can influence width and the rotation of the shake vector.By The light beam is shaken on the width in the groove, which produces in the coaxial direction Deflection in (along the groove), and there is negligible effect (cosine error e.g., < 0.2%) on the width.In the intersection PGC scaled error in axis directly influences groove width about 2% to about 3%.

However, very wide groove, which may have, is distorted its offset endpoint by the rotation, such as in larger raster pattern It is the same in case.Since this can influence the intersection of wide groove, PGC amendment is suitable for shaking.Each processing is segmented the distortion Item is transmitted to the DSP 1026 from the SCC 1010, and can be represented by eight (8) positions, and 12.5% error model is provided Enclose with 0.1% resolution ratio, by 1/1024 scaling after.

6.AOD calibration process

Figure 23 describes the AOD calibrating pattern according to an example with figure.The following process is relative to the galvanometer framework 2312 and AOD scale factor rotates 2310 to calibrate the AOD framework.Implement to turn for the TMtert in the scanning field immediate vicinity The calibration changed, wherein the scanning field distortion can be ignored.Notice that the rotation of the galvanometer framework 2312 may be for shake and light It is not focus for grid.As long as the AOD framework 2310 is orthogonal and is directed at the galvanometer framework 2312, be correctly aligned shake and Grating, because these operations are relative to the galvanometer framework 2312.

The process includes setting TMtert conversion to default (the equivalent nominal SF of matrix *), and the X stage is moved on to it At the center of traveling.Then, the FPGA1028 is loaded using four sight line grating patterns 2314.In order to be used for discussing, Figure 23 Show amplification (and rotation) version 2 316 of one of the sight line grating pattern 2318.Each pattern includes central sight line 2320 (zero AOD displacements), and the sight line 2322 for one of moving on to the AOD axis (+ch0 ,-ch0 ,+ch1 ,-ch1).Pay attention to these positions It moves in the original AOD coordinate framework (ch0,1) rather than in workpiece or galvanometer framework (XY).The process includes using to be somebody's turn to do The group that processing stand size aims at line pattern 2314 and/or each independent pattern 2318 to four zooms in and out, and can be based on should System setting is changed.The processing stand size is system configuration parameter.

The process also includes that each cuts to the grid chart centered on the scanning field center by four patterns 2314 In case.For each pattern, which is moved on at each position of the pattern, and solves (e.g., to reach before grating 1ms).In sample instance, all patterns can be fallen in the square from 1mm at the scanning field center.The process repeats this and cuts Cut four patterns 2314 each the step of up to pre-determined number, the mobile Y linear stage is to replace these patterns.For example, such as Shown in Figure 23, cleavable ten four patterns 2314 (or another pre-determined number based on the particular instance).The process includes Collecting data is to position replacement AOD sight line 2322 relative to their corresponding zero shift sight lines 2320.Then, should Process evaluation is somebody's turn to do (e.g., ten) data set, and calculates AOD scale factor (every micron of AOD MHz) and the opposite galvanometer coordinate structure The rotation of frame 2312.Notice that four center sight lines (zero AOD displacement) provide information relevant to galvanometer rotation.Ying Xiang The rotation of the AOD axis is calculated for the angle of the galvanometer axis.The process may include that verifying galvanometer pattern rotates less than Check about 1% when the optical setup.

I. scale factor interval and TMtert construction

The AOD calibration process generates two scale factors (SF):

MHzPerRaw μm: [AOD MHz]/[original galvanometer μm],

MHzPer μm: [AOD MHz]/[μm].

First SF (MHzPerRaw μm) has X and Y-component, constitutes the scaling item in TMtert conversion. As described above, third level section produces the AOD order in original galvanometer coordinate to coordinate to be somebody's turn to do during the third level positions Galvanometer subsystem 1044.Thus, which illustrates the combination of the AOD and galvanometer SF.

Second AOD SF (MHzPer μm) is by SCC shake (or grating) table data (with the unit of XY workpiece surface μm Point out) it is converted into AOD unit (MHz), before by the data download to the FPGA 1028.

The result of AOD calibration generates MHzPer μm.In order to generate MHzPerRaw μm, uses and be embedded in the galvanometer Galvanometer scale factor data in calibration.The galvanometer can be extracted from the XY scale factor that the galvanometer calibrates data SF (Raw μm Per μm=[original galvanometer μm]/[μm]), evaluated at the center of the scanning field of negligible field distortion. It is used subsequently to each of the XY component:

MHzPerRaw μm [X, Y]=MHzPer μm/Raw μm Per μm [X, Y].

In order to formed the TMtert conversion, by relative to the AOD framework of the galvanometer framework rotation with it is determined above Scale factor mutually merges.The AOD frame rotation can be it is non-orthogonal, because of the mechanical tolerance in the AOD1020, in 1022； It include thus two independent rotation items.Figure 24 describes the AOD collimation angle according to certain examples with figure.It can be by the AOD The rotation conversion of framework (TMframe) is defined as

In

Angle of the ThetaAod0=between ch0 the and X galvanometer axis,

Angle of the ThetaAod1=between ch1 the and Y galvanometer axis,

The amplitude scaling (μm AOD0/ μm) of K0=AOD ch0vs workpiece surface, and

The amplitude scaling (μm AOD1/ μm) of K1=AOD ch1vs workpiece surface.

As shown in Figure 24.

In Figure 24, in the central evaluation perspective and scaling of the scanning field.TMframe indicates to be used to the AOD coordinate structure Frame is aligned in the galvanometer coordinate framework.TMfield conversion amendment scanning field distortion and nominal galvanometer alignment, including this Any mark of galvanometer axis is overturn.In order to avoid obscuring, in some instances, pass through the rotation angle (additional 180 ° of rotations) Any mark overturning in the AOD is considered, does not pass through the K0/1 scale factor.K0 and K1 does not include this MHzPer μm scaling Factor.The scaling is used when loading and shaking (or grating) table, or is included in TMtert conversion (for profiling).Phase Instead, K0 and K1 illustrates the scaling variation of the possibility between AOD0 and AOD1, used the nominal MHzPer μm of scaling because After number.In some instances, K0 and K1 expectation is equal to 1.0.

Subsequent TMtert, the used conversion during third level section, provides as follows

Here, this MHzPerRaw μm [X, Y] scaling item should only include that amplitude scales；In the rotation item of the TMframe Consider label.

Mtert is used in instant computing, applied to the AOD order output at the third level filter:

It is noted that the definition of the TMtert means that rotation follows scale factor to correct.Thus, when the evaluation AOD standard When data, first TMframe conversion is used to for the AOD data being rotated into be aligned with the galvanometer framework, is evaluating the AOD Before deflection is to calculate MHzPerRaw μm [X, Y].

D. power control

The radium-shine function that power control is used for the LDA system to be consistent during shake and during speed changes Rate.Power control includes the linearisation of the AOD power attenuation, and the school of the linearisation AOD power control to workpiece surface power It is quasi-.

1.AOD power Linear

Figure 25 is the block diagram for power control signal stream described according to an example.Two linearisation tables 2510, 2512 be used to generate the ch1RF signal amplitude order 2514 with by the Optical output power linearly turn to RF signal frequency used and The function of required Optical output power.

For each of two AOD units 1020,1022 (ch0 and ch1), shake (e.g., is shaken 0 or shaken 1) The frequency departure (beam trajectory) order (Fdev0 or Fdev1) and nominal centre frequency (Fctr0 and Fctr1) are added to life At the entire frequency order (RfFreq0 or RfFreq1).Ch0 frequency order RfFreq0 indexes the ch0 linearisation table In 2510, the Pscale0 power scaling order is produced.The Pscale0 power scaling order is illustrated for keeping as frequency Rate function and the Optical output power scale factor of the ch0 Optical output power linearized.The linearisation table 2510 is only The one-dimensional functions of ch0 frequency.

Lai Fuyong Pscale0 (is issued and ordered) by the DSP 1026 by system power commands scale factor Kp, is formed Overall power scale factor Pscale.The order, along the ch1 frequency order RfFreq1, index enters two-dimensional look-up table 2512 to generate RfAmpCmd 2514, the ch1RF signal amplitude order.Note that the RF signal amplitude for being used for ch0 is kept not Become.In other words, all output powers in this sample instance are implemented by the RF signal amplitude modulation of ch1.

I. table is linearized

The AOD 1020,1022 controls optical power by changing the RF signal power level for the AOD unit.Figure 26 describe example power controlling curve (the RF letter of the Optical output power vs. standardization of standardization according to an example with figure Number power).Notice that Figure 26 describes the optical power output for being formed in RF amplitude order.

In some instances, by Figure 26 nonlinear curve mapping with generate by required standardization output power (from 0 to 1,1 is peak power output) it is converted to the AOD RF signal function for needing to be used to obtain the output.This can be taken as to the X for Figure 26 With the overturning of Y-axis: providing required optical power, determine the required RF amplitude order.

The linearization curve of the formation is shown in FIG. 27.Figure 27 describes the example ch1AOD according to an example with figure Power linear curve set.Note that the power control profile in Figure 26 illustrates the AOD behavior at single RF signal frequency. In fact, the curve changes with the RF signal frequency for the AOD unit 1020,1022.Thus, for the AOD subsystem Several RF signal frequencies on 1042 operational frequency range repeat the Nonlinear Mapping.Figure 27 shows the linearization curve Collection.

The linearisation table generates to generate the required RF amplitude order of the requested Optical output power scaling.The RF Signal amplitude output affects the Optical output power, but the output power is not arranged directly.

In contrast to the ch1 table, which produces the Pscale0 power scaling factor (rather than RF amplitude order). The ch0 is responded the function for linearly turning to ch0RF signal frequency by Pscale0.The ch0RF signal amplitude is left intact and does not have to In the linearisation table.Figure 28 describes the example plot for ch0AOD according to an example.Figure 28 is shown in ch0 optics Variation vs RF signal frequency (left figure) in efficiency, is formed in the required power scaling for linearisation (right figure).

As shown in Figure 28, which is linearized in their frequency ranges, is made by reducing The output power at optional frequency is made to be equal to the power at the minimum efficiency frequency for the amplitude order of frequency function.Cause And the effective optical efficiency for reducing the AOD subsystem 1042 is linearized, and the AOD 1020,1022 operational frequency range Influence with the optical efficiency to the subsystem.

Ii. high-power mode: efficiency gain

Above-mentioned power Linear process is capable of providing the abundant power line in whole operation RF signal frequency range Property AOD subsystem 1042, be equal in ch0 and ch1 the product of worst condition optical efficiency the optical efficiency that successively decreases.It can The AOD subsystem 1042 is conservatively operated in this mode, is saturated using correct arized power and without any RF amplitude order.

However, can be by calculating as ch0 the and ch1RF signal frequency range (or, being equal to, AOD XY deflection Range) 1042 optical efficiency of practical AOD subsystem of function realize significant optical efficiency gain.This allows the AOD Subsystem 1042 operates in than calibration of power range narrower range entire at higher-wattage grade, does not make the RF amplitude Order saturation, because high high light efficiency needs lower RF amplitude order.Or, equally, this mode allows the AOD 1020,1022 carry out the calibration of power on than normal very big frequency range with big grid fields or very wide groove, and The normal operating in the more typical range of deflection less than normal is not influenced.

In one example, it calculates the optical efficiency gain to pass through, determines and (be used for ch0 and ch1 bis- in selected frequency range Person) on minimum efficiency, calculate the efficiency of the worst condition by the value multiplied by two worst conditions, and be several possibility Frequency range and repetition.This process results the conservative curve for efficiency gain vs. frequency range, be suitable for general service and When carrying out grating.It is possible for further improving when considering vector processing, wherein dividing between shake and AOD transient state AOD deflection.Shake produces the AOD deflection along the line with angle rotations some in the AOD coordinate component.For this Practical AOD range in ch0 and ch1 deflector is based on the shake angle (such as sine and cosine functions).Any axis is all not simultaneously The entire deflection of experience.Thus, which could possibly be higher than the conservative estimation of the above generation.

However, it is also possible to consider AOD transient excursion during shake.Such AOD transient excursion is likely to occur in any Axis is based on the beam trajectory.

Figure 29 and Figure 30 is the figure for describing the efficiency gain calculated result.Figure 29 is the example described according to certain examples The figure of AOD efficiency curve.Figure 30 is the figure for describing the example A OD efficiency gain according to certain examples.For example, Figure 29 shows two The efficiency of a AOD unit 1020,1022 (ch0 and ch1), and Figure 30 show for grating and example both of which relative to This entirely linearizes, the efficiency gain of unsaturated efficiency.

The 2.AOD workpiece surface calibration of power

After the completion of the above linearization procedure, which is loaded into the FPGA 1028AOD controller.It sends Standardization power command (Kp, range is from 0 to 1) to the FPGA 1028 has subsequently formed and can be the order power magnitude The optical power of linearized function exports, and independently of the RF frequency order for being used for the AOD 1020,1022.In certain examples In, the quantization of linearisation table and calibration uncertainty produce the residual linear error between about 1% and about 2%.

However, in some instances, the scaling of arized power output is arbitrary.In last calibration of power step In, several values between zero and one are set by the radium-shine power command of the standardization and chuck power table (CPM) measures the shape At workpiece surface power.Linear fit has determined CPM offset and scale factor Kpwr, by workpiece surface power (watt) It is converted to the standardization power command Kp for being sent to the FPGA 1028 from the DSP 1026 during processing.

Figure 31 is the block diagram for describing the calibration of power data process 3100 according to an example.The power shown in Figure 31 Control data process 3100 also supports the high-power mode enabled by efficiency gain.This includes the x2 in the FPGA 1028 Gain makes it possible for the efficiency gain greater than 1 to amplify part Kp value.Additional 1028 scale factor KpNorm of FPGA To the Kp ordering norms, so that Kp=0.5 illustrates the unsaturated linearisation output power of the maximum, and it is greater than 0.5 value It illustrates because efficiency gain may acquire (unsaturation) higher-wattage within the scope of the AOD of reduction.

A method of the workpiece surface calibration of power is provided, according to an example, including implements the AOD power Linear school It is quasi-.According to the data, which determines PscaleCal, minimum ch1 efficiency.This method further comprises carrying linearisation table Enter the FPGA 1028 and sets PscaleCal for the 1028 scale factor KpNorm of FPGA.Then, this method include should 1026 power scaling factor Kpwr of DSP is set as 0.5 and issues the power command equal to 1.In view of inside FPGA 1028 Scaling, which is limited to the ch1Pscale value to be no more than PscaleCal, in the power Linear calibration range Any Fdev0 or Fdev1 frequency order.Which ensure that can get the linearisation in unsaturated entire AOD frequency range Output power.Notice when Pscale0 (ch0 linearize table output less than 1 when, Pscale is likely less than PscaleCal, But this is occurred only at the ch0 frequency with high light efficiency.Thus the practical Optical output power is nominally in institute Have retain in Fdev0 frequency it is constant.The same principle is also effective to ch1 linearisation output.This method also includes record LinPmax, when measurement (being set) in previous steps using the entire linearisation, unsaturated power command by the chuck power table Workpiece surface power.Then, 0.5/LinPmax is set by the 1026 power scaling factor Kpwr of DSP.This standardization 1026 power command of DSP (being indicated with watt) makes 1028 power control of FPGA produce the calibrated workpiece surface function Rate.The power command (being indicated with watt) may have exceeded the LinPmax for AOD range now, and wherein the efficiency gain is big In 1.In such a case, Pscale0 and/or ch1 linearisation table entry is set smaller than 1, so that the width of the formation Value order keeps unsaturated.

In short, the following information can be deposited as calibration data, and the DSP 1026 and FPGA is loaded into during initialization 1028:Ch0 linearizes table；Ch1 linearizes table；KpNorm；Kpwr；Efficiency gain table；And LinPmax.

E. process velocity limits

For groove, determine that the maximum processes segmentation speed by several factors, comprising: shake timing (needs a little big It is small to overlap between shake row)；Because of the AOD dynamic of third level filtering；For processing radium-shine power obtained by the groove；With Data rate limitation.

1. buffeting speed limits

Because the rate limitation of shake is formed in the overlapping being used between shake row.In some instances, in shake row Between maximum position increment be 0.35*Deff (it is assumed that 65% overlap to provide satisfied margin).Thus, for keeping this The rate limitation of overlapping is 0.35*Deff/Td=0.35*Deff/ (Nd*Taod).

On the time of each shake row, dosage and width parameter are remained unchanged, has and only goes in next shake Any variation for working of section start.Since 1028 control algolithm of inferior grade FPGA implements these, by 1026 institute of DSP The parameter for being computed and transmitted to the FPGA 1028 can be by linearly interpolation；This is updated in right times by the FPGA 1028 to tremble Dynamic row parameter.

Generally, it designs shake discussed above to generate so that shake update is not restrictive factor, the system is allowed to exist It is run at the limitation of radium-shine power.

2. third level filter limits

The processing of the third level filtering of calibrated AOD sizes of limitation is given, limitation is added on to process velocity.

During third level filtering, the Spline smoothing in speed between processing segmentation is produced in the AOD order Transient state respond.It is responded for example, Figure 32 is described with figure according to the third level filter to velocity variations of an example.The response Amplitude it is proportional to the Spline smoothing of the speed, and the die-away time is the letter of the third level filter bandwidht and damping ratio Number.

Figure 32 shows the VELOCITY DISTRIBUTION 3210 for maximizing the worst condition of AOD offset, occurs when a speed When variation 3210 (being equal to 2*Vmax) is followed by second speed variation 3214, there is equal amplitude but opposite label, timing In the overshoot 3216 of the third level filter (for the third level filter of 3kHz, about 0.12 millisecond after speed change) Peak.

If AOD transient state scale factor is defined as " Ktrans ", deltaV is changed for processing segmentation speed,

DeltaAod=Ktrans*deltaV.

The example value of Ktrans is 26.6 μm/(m/sec), for the third level filter of fourth order 3kHz.Thus, example Such as, for can make+2 to -2m/sec velocity variations the process velocity with 2m/sec segmentation, about this Limitation=2* (2m/s) * (26.6 μm/(m/s)) of deltaAod=106.4 μm.

Figure 33 describes the Monte-Carlo AOD Transient according to an example with figure.With random sequence and at random It is above-mentioned opposite that speed is segmented the single MonteCarlo model validation of the random velocity sequence 3308 of amplitude (reaching+- Vmax) In the scene restriction of Figure 32, the AOD deviates 3310 (as shown in figure 33).Therefore third level filter response is reliably predicted The AOD offset of worst condition for giving radium-shine process velocity.

3. radium-shine power limit

Basic limitation is added on process velocity by radium-shine power.For the first rank estimated value, need for processing groove Dosage (power/speed=W/ (m/sec)=J/m) is based on the groove area.For the groove cut by shake Gaussian beam, The area of the cross section is about 0.65* width * Deff, in the wherein sample instance of Deff=available point size.When logical When crossing shake increase width, which is D* (0.65*Deff+ width-Deff).

The example model of volume requirements is

Dosage (J/m)=Area (μm 2)/143+0.3).

In calibrated LDA system, which provides the calibrated value for dosage, wherein

Dosage=Fluence*EffectiveWidth.

Since dosage is equal to power/speed, which has been determined for the given maximum speed that can get radium-shine power Degree.

Figure 34 describes the example rate limitation with Taod=Tcmd=1 μ sec according to an example with figure.Figure 34 packet The example plot of the example plot for dosage limitation and point (Npt) limitation for variable number is included.The workpiece surface radium It penetrates power limit (about 8W is used for example LDA system) and limits the process velocity of the function as width, as shown in figure 34.Root It may be optimism according to the example plot that certain examples describe, because they do not include the AOD efficiency (e.g., in the maximum wideband 65-70%) or for additional AOD needed for the third level section deflection effect.Nevertheless, about radium-shine limitation process velocity The upper limit constitute conservative lower limit, by other rate limitations (beam positioner and shake) be more than to avoid limit The quantum of output of property processed.The dither table of maximum quantity point (max Npt limit) has been used to need maximum width if Figure 34 is highlighted Degree, the fact that by the maximum speed at smaller width is unacceptably limited in.

In fact, maximum workpiece surface power is the function of optical series efficiency, change with AOD range (due to this AOD efficiency curve).The required groove width has determined the mesh width (required shake deflection), and that define maximum powers Grade (because AOD power Linear).Third level AOD deflection is also included in the calculating, because it needs additional AOD inclined Turn and thus lower maximal efficiency.Thus, obtainable radium-shine to be reduced with groove width and influence the curve in Figure 34.

The required AOD range is provided as the function of segmentation speed Vseg

AODrange=[gridWidth+4*Vseg*Ktrans] * 1.10.

This include component due to: the required mesh width for groove is (if the maximum mesh width can become Change)；Because the dynamic AOD offset (passing through " Ktrans " scale factor) of third level filter (pays attention to for segmentation speed Vseg, maximum speed variation is 2*Vseg, and the pk-pk AOD range of deflection is 2* (2*Vseg) * Ktrans)；With 10% Be satisfied with margin, including scanning field distortion and shake angle effect based on speed.

Sample result is shown in FIG. 35, and describes the AOD stroke range vs. characteristic width according to an example with figure. The example in Figure 35 has used 2m/ seconds speed, the radium-shine power of 8W and 280 μm of AOD range.The mode also predicts Than the maximum grating diameter that can be processed without being disposed after the movement to the stop position.This may be example μm such as, < 100-not sufficiently large for most of pad but sometimes enough for trench intersection.Larger grating The alternative in region is that placement reaches the period defined in third level filter response (generally at the optical grating point 0.4-0.5 milliseconds).

Note that Figure 35 does not include the effect of the AOD range about AOD efficiency.In order to engage to process velocity calculating It influences, the AOD range of the calculating is merged with the efficiency curve (Figure 30) to calculate obtainable radium-shine power as Vseg's Function.It is then able to the radium-shine power limit determined on process velocity, it is (fast based on relying on as the radium-shine power by obtained by The AOD range of degree) compared with required radium-shine power (based on record).For example, Figure 36 describes adding according to an example with figure Radium-shine power limit in work speed.In the example of Figure 36,50 μm of groove is shown, wherein the process velocity is 1.65m/ sec。

I. arc machining effect

During wide circular arc, which is amplified by (1+W/R/2), the lower center line in the circular arc can be promoted Speed is to avoid the power limit in the outer rim.But the above mode shows any step being suitable in speed and (is less than Twice of the datum speed) variation, as long as the process velocity for having selected this nominal is to adapt to the track AOD of worst condition.Thus Wide circular arc can be processed and is not concerned with the limitation of AOD stroke range.

4. data rate limits

Data stream rate limit be provided with to this be subjected to process time it is relevant compared with lower boundary for any processing segmentation (about 7 μ sec).The length of processing segmentation is provided, this is provided with the upper limit relevant to the segmentation speed.For example, 21 μm of segmentations To there is the rate limitation because of the 3m/sec of data rate.

5. process velocity limitation is summarized

In one example, for the calculating of the process velocity of groove according to these steps.Technical staff will disclose from there Content recognize can be certain in the following steps to implement with different order.

In the first step, using the manuscript to determine Fluence based on groove dimensions (width and depth), GridWidth and EffectiveWidth.

In the second step, Dosage=Fluence*EffectiveWidth is calculated for the groove.The required radium-shine function Rate is then Preq=Dosage*Vseg as the function of process velocity Vseg.

In third step, the function of the required AOD range of deflection as Vseg is calculated.Using the result with determine this by In the rate limitation of AOD stroke range (Vaod), the calibrated AOD range of the maximum is provided.

In four steps, using the efficiency gain curve (see " high-power mode "), the letter as AOD range is calculated Radium-shine power obtained by several maximums.

In the 5th step, which is determined according to radium-shine power (Vlaser): where the required radium-shine power Process velocity be equal to the obtainable radium-shine power.

In the 6th step, (Vdither) is overlapped according to shake to determine the rate limitation.

In the 7th step, which is determined according to minimum split time (Vsegtime).

In the 8th step, Vaod, Vlaser are set by the process velocity, the minimum in Vdither and Vsegtime Value.

IV. cooked mode

In one example, the LDA system is with three kinds of different mode rapidoprints.Vector pattern process variations width and depth Linearisation track, arbitrarily the two can be controlled.The processing of grating mode creates the smaller feature of any 2D shape, In an AOD scan field (e.g., ten times of point sizes greater than the processing).Scrape mode processes the big region of arbitrary shape, tool There is the good control in the precision of the peripheral position and the scraping region in the consistency of the ablation depth of material.

A. vector is processed

The vector processing of groove is shaken dependent on AOD to control groove width.In certain examples of LDA system, for example, There may be to shaken and the relevant focus of inconsistent mode, wherein the dither pattern originates in the beginning in the groove With any position at end point, and have the shake angle that changes with path velocity (relative to the main light beam rail Mark).

In order to handle the focus, an example is that the integer row of each processing segmentation takes shake to generate unanimously , repeatable and suitable values segment end.Processing fragmentation command can be cut out to support such integral jitter Capable timing.

Multiple integral jitter row supports the track-end transition region for being used to generate crosspoint.Exemplary 40-50 μm turns The depth tolerances changed 2-3 μm of error for allowing light beam to position and retain 5-10%.For the example used in the LDA system 25 μm nominal of available point size, the transition length are converted using the dosage of 4-5 shake row.

Dosage is remained unchanged to be retained in the not Trapezoidal at trace end between each shakes the departure date.It can be based on every One shake row is changed to be created the predictable groove width during doses change using the control of all dosage.

In some instances, use shake so that continuous point by the spot diameter > 60% overlap.It can also make With shake enough near the overlapping for being capable of providing between continuous row > 60%.For wide groove, this is likely to become a class Topic, because the Vibrating pulse of required big quantity, and the constraint to the AOD more new period.

The AOD more new period is constrained by the conversion time of the acoustic wave in the AOD on optical light beam.The LDA system Unite A0D 1020,1022, in one example, can be operated with the update times of > 500 nanoseconds.With quick renewal rate, because Transient state in the acoustic wave and the distortion in the inflexion point becomes very noticeable；Even if point distortion may with 500 nanoseconds It is problem.Spot diameter amplifies the fluence for reducing workpiece surface, leads to change in depth.In one example, to be faster than for 750 nanoseconds Rate updates the AOD 1020,1022, when being the practical conversion of the acoustic wave in the AOD on 4.5 millimeters of optical light beams Between.Figure 37 is described with figure and is distorted according to the example of certain examples.From left to right, Figure 37 will in 440 nanoseconds, 640 nanoseconds and The distortion of two inflexion points at 1240 update cycles nanosecond is compared.

Another example of the LDA system is run with very high beam trajectory speed.The fair speed, point overlapping demand and The combination of limited AOD renewal rate means that the jitter points quantity changes with this feature width.In other instances, it uses One dither table and change this feature size by scaling the distance between the jitter points.

1. jitter parameter calculates

In some instances, which defines the calculating and foundation of jitter parameter.

Input parameter:

Deff: available point size (μm)；In the width of the non-wobbling groove of the required depth.Pay attention to Deff with the optical point Diameter change, together with the required depth (because dosage growth needs to increase depth and is also required to increase width).

BiteSizeRatio: the every maximum deflection sufficiently to overlap to keep smooth fluence to be distributed is provided, is expressed as a little Diameter score.BiteSizeRatio is≤0.4 for Gauss point.

Taodmin: minimum AOD update cycle (μ sec)；400-700 nanosecond is surrounded for an example of the LDA system.

The Tclock:AOD clock cycle (μ sec)；The period is arrived into Taod quantization.Nominally 0.04 μ sec.

Output parameter:

Taod: practical AOD update cycle (μ sec).

Nd: for jitter points quantity needed for required groove width；The function of width and Deff.

Td: entire shake period (μ sec)=Nd*Taod.Machined parameters (width, depth, shape, slope), for each A shake row remains unchanged, and is updated with the rate.

Tcmd: the time cycle (μ sec) between the update of the light-beam position nominal to this.Nominally 1 μ sec.

Ncmd: to the Tcmd command cycle between the variation for processing variable and/or shake parameter (Nd, Taod) Quantity.

It is borderline in Tcmd to update due to there may be the jitter points of variable number, can adjust the AOD update cycle Entire row.In addition, quantifying the AOD update cycle to 20 nanosecond intervals.

The calculating of the dependence parameter starts from the required characteristic width.The width of the shake track is estimated as by stringent

Width=Deff+Nd*Deff*BiteSizeRatio.

It rounds the jitter points quantity for possessing the required width up discrete value, and is given

Nd >=1+ceil ((Width-Deff)/(Deff*BiteSizeRatio)),

Wherein " ceil " is the upper limit function for rounding next integer up.

Because rounding up, needed for Nd may be generally greater than, and BiteSizeRatio can be reduced so that the width is zoomed to this Desirable value (passes through the Kw parameter).The command parameter is given by the value for providing Nd

Ncmd=ceil (Nd*Taodmin/Tcmd), and

Taod=ceil (Ncmd/Nd*Tcmd/Tclock) * Tclock.

For example, for Deff=25 μm, width=250 μm, BiteSizeRatio=0.4, Taod=0.5 μ sec, and Tcmd=1 μ sec,

Nd=ceil ((250-25)/(25*0.4))=23,

Ncmd=ceil (23*0.5/1.0)=13, and

Taod=ceil (13/23*1.0/0.02) * 0.02=0.580 μ sec.

The maximum deflection range of the AOD 1020,1022 of the LDA system is equal to 10*Deff in one example.If used The entire AOD range, Nd≤24.

Above-mentioned equation gives the general solution of jitter parameter.May with greater need for be to maintain Taod constant and implement Nd's Value, which create the discrete values of Ncmd.For example, may be 2 power and Taod=0.5 μ sec for Tcmd=1 μ sec, Nd, Or Nd may be 4 power and Taod=0.75 μ sec.As discussed in the next part, the AOD frequency limit, light It learns point quality and above-mentioned equation collectively forms the LDA system, according to an example, used 1 μ for Taod and Tcmd The sec period.

2. nominal frequency

Shake is used as relative to nominal AOD frequency to the frequency departure of (Fnom0/1), indicates the nominal light Beam path (that is, center line of the groove).With renewal rate (Tcmd seconds every) Lai Gengxin Fnom0/1 of the order, and thus one As every shake row update it is multiple.In one example, it is completed immediately using the Fnom0/1 without waiting for shake row, for keeping Precision in the general beam direction track.The comparison of the update of the machined parameters (width, shape, dosage, slope), each It is remained unchanged on a shake row.

However, Fnom updates, in some instances, (each at next obtainable AOD update is occurred only at Taod seconds).If Fnom update is asynchronous in the AOD update cycle, they have effective temporal uncertainty equal to Taod. In addition, then the amplitude of the vibration changes if Taod is not remained unchanged and (e.g., calculated the function as Nd, as described above).

The precision that processing is reduced by the introduced uncertainty of nonsynchronous operation, because the random vibration cannot be by school Standard is fallen.For example, for Taod=750 nanosecond, it is uncertain with ± 0.75 μm of additional position with the groove of 2m/sec processing.

In order to avoid the error, Tcmd may be the multiple of Taod.It is assumed that each DSP1 servo period (5 μ sec) will more It is newly sent to the FPGA 1028, the original value with Tcmd=1 μ sec, the Taod was 500 or 1000 nanoseconds.Another choosing Selecting is Taod=Tcmd=750 nanosecond, although the data packet of this variable number that may need each servo period.

As shown in figure 37, the distortion of the deflection light beam except the AOD is in about 500 critical nanoseconds.In order to keep away Exempt from the distortion, example modes processing, which uses, to be run Tcmd=Taod=1000 nanosecond.In additional examples, AOD design may Allow the smaller value (e.g., 250 nanosecond) of Taod.Thus, the variation of Structure adaptation Tcmd and Taod.

3. shaking angle modification and amplitude scaling

In one example, by the orientation adjustment of the AOD dither pattern to keep its direction orthogonal (that is, vertical) nominal in this Beam trajectory.For example, Figure 38 describes the shake perpendicular to nominal beam track 3812 of the alignment according to an example with figure Row 3810.In view of using systematic nomenclature described in Figure 38 because the processing of the offset of beam velocity is described as follows.It will Each radium-shine position 3814 (being shown as 11) is aligned and ignores the speed of the radium-shine light beam linearly along the shake row 3810 Degree, when it is mobile relative to the beam trajectory 3812.

I. label conversion

It is the same in sample instance as disclosed herein, the light beam definition of shake is positive shake in the galvanometer X-axis, And it is then rotated in the amendment direction to keep the beam orthogonal of the shake in the beam trajectory 3812.By the AOD subsystem 1042 calibration of system is to match the galvanometer XY coordinate framework in the scanning field center.The following equations have studied shake rotation Torque battle array.Unless otherwise noted, it is positive (CCW) counterclockwise for the convention of angle, has along the zero of the+X galvanometer axis Value.

Ii.AOD rotation and transition components

Usable following equations calculate the final AOD order:

AODcmd=TMframe*TMfield*Rdither*Dither,

Or

AODcmd=TMaod*Dither,

Wherein:

AOD RF frequency order (2- component vector) of the AODcmd=to each RF driver；

AOD XY order is transformed into the nonopiate transition matrix of the AOD coordinate framework by TMframe=, such as in the school AOD It is the same defined in quasi- part；

TMfield=locally scanning field distortion amendment, in galvanometer XY framework (4- module matrix)；

Rdither=is by matrix rotation to determine the shake vector relative to the beam trajectory 3812 (4- module matrix) To；

Dither=is loaded in shake (or grating) table in FPGA 1028；With

The complete AOD order transition matrix of TMaod=forms from the said modules-component of the matrix being converted into this FPGA 1028, replaces the Kw0, and Kw1.

Using the TMfield correction term (two-axis scaling and rotation) to consider local scanning field distortion.This can be used because Arbitrary scan field calibration amendment is just being operated without for the AOD 1020,1022.The conversion may originate from galvanometer calibration money Material.The component of the matrix is

Wherein:

SFx=X- axis scaling (nominal 1.0)；

SFy=Y- axis scaling；

The Ryx=Y-axis is to the rotation in the X-axis (rad)；With

The Rxy=X-axis is to the rotation in the Y-axis.

The origin iii.Rdither

The light beam of the shake has coaxial and intersecting axle component, has amplitude (on a shake period) are as follows:

DitherCA=Width,

DitherOA=Vel*Taod* (Nd-1).

Notice that the OA jittering component " backward " makes relative to the light beam of the mobile shake of the beam trajectory 3812 every The coaxial position of all the points in one shake row is kept fixed that (e.g., the radium-shine position 3814 is along the shake row in Figure 38 3810 are kept fixed).Thus, which is based on the velocity vector to keep each shake row perpendicular to the light beam rail Mark 3812.OA the and CA jittering component merges to form " shake vector " with the shake angle relative to the+X galvanometer axis. As shown in Figure 38, which defines velocity compensation angle, θ_velAre as follows:

θ_vel=atan (DitherOA/DitherCA).

Shake vector orientation includes the velocity compensation angle, θ_vel, in addition trajectory angle θ_trajWith 90 ° of rotations (that is, should 90 ° are described in the beam trajectory 3812 and Figure 38 between OA jittering component DitherCA):

θ/2 θ dither=θ vel+ θ traj+.

Thus, the shake angle, θ_ditherIt is the entire shake vector angle relative to galvanometer XY framework.Then pass through The angle, θ_aodThe shake vector is aligned in the AOD framework, may be laid out according to the optical series relative to the galvanometer XY framework rotates.

During processing, the immediate updating trajectory angle (θ_traj) (the letter as beam trajectory velocity component Vx and Vy Number), and the AOD angle (θ_aod) constant after calibration.For the immediate updating (1 μ sec update), trigon meter can be minimized It calculates.This can be realized by using trigonometric identity: such as, sin (atan (y/x))=y/sqrt (x²+y²)。

Can by the DSP 1026 using the following parameter to calculate jitter direction:

The entire scope of ditherRange=current dither table (μm)；

Kw=creates the required scale factor for shaking width for the dither table, and (this can be interpolated for taper point Section)；

Jitter points quantity (for each segmentation constant) of the Nd=in current dither table；

Taod=dither table renewal rate (μ sec)；

The X and Y-component of the beam trajectory 3812 of Vx, Vy=in original galvanometer XY coordinate (m/sec)；With

Vel=sqrt (Vx²+Vy²)=velocity vector amplitude.

The subsequent trigonometric identity allows the sine of the Rdither spin matrix and cosine term that will be passed through by calculating

// nominal intersection dither axis width after scaling

DitherCA=ditherRange*Kw (μm)；

// one shake the departure date between to keep shake point alignment required coaxial increment

DitherOA=Vel*Taod* (Nd-1)；

// according to the amplitude (vector sum of OA, CA) of the full shake vector of velocity compensation

DitherMag=sqrt (DitherOA^2+DitherCA^2)；

// avoid being divided by zero result；It is silently identity matrix

if abs(Vel*ditherMag)<1e-6

CosThetaDither=1；

SinThetaDither=0；

else

CosThetaDither=- (Vy*DitherCA+Vx*DitherOA)/(Vel*DitherMag)；

SinThetaDither=(Vx*DitherCA-Vy*DitherOA)/(Vel*DitherMag)；

end。

Iv. scaling and TMaod are calculated

The amplitude of the shake vector be can adjust to consider the additional the length of the hypotenuse in the shake vector of the erection rate. Thus,

KwCorr=Kw*DitherMag/DitherCA=DitherMag/DitherRange.

The last transition matrix is then provided by the matrix product

Two (TMframe, TMfield) before this can be precalculated in the SCC 1010 and be transmitted in this In " TMdither " transition matrix:

Can instant computing subsequent two (Rdither, the KwCorr), the Kw based on the velocity vector and interpolation.Pay attention to Rdither and KwCorr includes the RditherMag, and product Rdither*KwCorr causes RditherMag to be taken The calculating of the sqrt () for the Expenses Cost for disappearing-avoiding in the DSP 1026.If we define

CosThetaRditherCorr=- (Vy*DitherCA+Vx*DitherOA)/(Vel*RditherRange)；

SinThetaRditherCorr=(Vx*DitherCA-Vy*DitherOA)/(Vel*RditherRange)；

Or, alternatively (calculating for avoiding DitherOA and DitherCA),

CosThetaRditherCorr=- (Vy*Kw/Vel+Vx* (Nd-1) * Taod/RditherRange)；

SinThetaRditherCorr=(Vx*Kw/Vel-Vy* (Nd-1) * Taod/RditherRange)；

Then

Four components of the TMaod are then transmitted to the FPGA 1028, the update of RF signal frequency is calculated as

Wherein shake X and shake Y is the entrance in the AOD dither table.Grating is calculated according to identical processing.Institute as above It states, the RditherY entrance is zero in vector or scrape mode by convention；RditherX is filled by the RditherMap To generate the shake entrance.

These in the TMframe matrix are constants and can be pre-calculated and be stored in the DSP 1026.It will be at this These in TMfield matrix are proposed from the galvanometer calibration data.The conversion is used for galvanometer XY framework, because The TMframe rotates possible larger (for compared with high rotation angle degree, rotation is non-swappable).

V. additional calibration TMaod is calculated

Certain examples include the calculating of the TMaod during shake, and that takes into account the calibration corrections to the Track command Effect.In the above-mentioned derivation showed, it is assumed that ideal (workpiece surface) speed.However, in some instances, will scan Field distortion amendment is used for the Track command, before transferring them to the DSP 1026.These have modified track speed Degree, as calculated as the DSP 1026.This create the minimal error in the shake vector speed angle calculation, It may cause a few micrometers of error.In order to avoid further correcting, it is seen that local field distortion amendment is embedded in the TMdither square Battle array.

Firstly, redefine form rotation and the transition matrix that scales the shake vector these:,

Wherein it is defined as by intersecting axle and with dither axis coefficient of angularity

Kca=Kw/VelIdeal, and

Koa=Taod* (Ndither-1)/ditherRange.

Both Kca and Koa can be calculated by the SCC based on processing segmentation parameter, using in ideal coordinate Path velocity.

The statement of the formulation presents various embodiment selections.In one example, by the value of Kw in processing segmentation It remains unchanged, and single constant TMdither is converted from the SCC 1010 and is transmitted to the DSP1026.Keep Kw it is constant may be Tolerable limitation in certain examples, because the line of taper can be created by short fragment sequence, each segmentation has Constant Kw value.In another example, the item for being used for TMdither is extended to create two conversions a: constant and one It is zoomed in and out by Kw.The DSP 1026 is then able to update the TMdither in processing segmentation:

TMdither=TMdither1+Kw*TMdither2.

Then it is capable of the value of interpolation Kw in processing segmentation to create taper line width.This means that being carried out by the DSP1026 More calculating, and from the SCC 1010 to DSP 1026 transmit additional information.The tradeoff is to provide shorter and more accurate wide Spend the ability of conversion.

The final shake order is generated by the FPGA 1028:

But by convention, only the DitherX component is non-zero in the FPGA dither table.Thus during shake It is only necessary to the first rows of the TMaod:

Based on the equation in the preceding section, which is given

However, in the DSP 1026 obtainable beam trajectory data be in original galvanometer, rather than it is ideal Coordinate.By the angle and amplitude of the beam trajectory velocity vector that scanning field distortion is modified, it is distorted the shake vector.In order to The shake vector in ideal coordinate is correctly calculated, (can be retouched local scanning field distortion by TMfield according to certain examples State) it is removed from the velocity vector, before calculating the shake vector:

Convention permission calculates the shake vector in ideal coordinate, is then converted into original AOD coordinate.The process Used three kinds of conversions: TMfield (original galvanometer is to preferably), TMfield (preferably arrives original galvanometer), and TMframe (original galvanometer to original AOD).Then the first row of TMaod is given

Or

It is that each segmentation calculates the TMdither matrix, and sends it to the DSP 1026 use by the SCC 1010 In the processing based on VxRaw and VyRaw instantaneous value.This then simplifies DSP processing, because these shake inverted terms are in the SCC It is pre-calculated on 1010.

The entire TMaod matrix for being transferred to the FPGA 1028 from the DSP will be then

This form of TMaod is used only for the special of shake.Grating is processed, entire TMaod matrix is generally used.

Vi. jitter direction controls

The direction (relative to the speed trajectory) of shake can be used, it is e.g., biggish when consistently processing during scraping When ablated area.Experience have shown that going out, the radium-shine machine direction relative to detritus secondary air can influence processing result significantly. Generally, processing " into wind " is preferably used to avoid interacting with the detritus plumage cigarette.

Above-mentioned equation is provided, shake can be simply changed by changing the label of the Kca in the TMdither equation Direction effectively overturns the shake movement of the intersecting axle.Can the SCC grade implement these, the speed based on the scraping access Spend track.In the row for paying attention to the regular distribution that generally scraping access is arranged in an angle direction, enable opposite In the simple control of the jitter direction of the detritus secondary air.

4. shaking row dose control

In some instances, the agent that the embodiment of the dose control during shake is remained unchanged in each shake row Amount.This allows dosage oblique deascension in the end of processing segmentation, and is maintained at the dose uniformity on the width of the line.This may have For for example, having the fast dose slope of wide shake line.

Figure 52 describes the shake row dose control according to an example with figure.The top graph 5210, which is shown, is distributed in two A series of shake rows 5212 (showing six) of processing segmentation (be shown as segmentation N and be segmented N+1, separated by dotted line 5214).The Two Figure 52 16 show coaxial (OA) shake order 5217 of each shake row 5212.Third Figure 52 18 shows each Intersecting axle (CA) the shake order 5220 of a shake row 5212.Notice that the OA jittering component has insignificant slope, keeping should Shake is stablized relative to the workpiece, when the general beam direction track is mobile with regular speed.The bottom diagram 5222 shows the reason The fluence order 5224 and practical fluence order 5226 thought.Pay attention in the second processing segmentation (left side of the dotted line 5214), The fluence oblique deascension (e.g., for form a cross street).The slope for continuing fluence is converted to a series of fluence step, With the fluence remained unchanged for each shake row.Each is provided by the fluence ramp value at the shake row center to tremble The fluence grade of dynamic row.

It is also noted that the coaxial position of the shake row 5212 not the beginning and end with these processing segmentations to straight, With the deviation for being equal to Taod*Vel* (Ndither-1)/2.This maintains the processing fractional concepts of broad sense, make it possible in office Meaning direction processes any segmentation.Any fabricated product as caused by the small deviation can be alleviated, by reducing in the spy Speed in fixed segmentation is to reduce shake row interval.

There has been described the modes for keeping each shake row of fluence constant.If allowing to shake width in processing segmentation When change, identical mode can be used.

5.TMaod limitation

In some instances, the amplitude of the TMaod entrance is limited to < 2.This can see in following derivation:

| cosThDitherCorr |=(Vy/Vel*DitherCA+Vx/Vel*DitherOA)/DitherRange,

Or

| cosThDitherCorr |=sinThVel*DitherCA/DitherRange+cosThVel*DitherOA/ DitherRange。

For holding point overlapping, the interval for shaking table entry is less than~0.35*Deff.In the same manner, segmentation speed is limited So that the interval between the shake row (ditherOA) is less than~0.35*Deff.Which ensure that DitherOA/DitherRange < =1, even for worst jitter conditions (Nd=2).By limiting, DitherCA/DitherRange=Kw≤1.Therefore | CosThDitherCorr | maximum value occur these each=1 when, and

| cosThDitherCorr |=sinThVel+cosThVel,

It has the maximum value 1.414 at ThVel=45deg.Identical limitation is used for | sinThDitherCorr |.

TMfield these may have less times greater than 1 (such as, after some time it is possible to reach 1.2) amplitude because scanning lens loses Very local scale factor.The item that nonorthogonality can create in TMframe reaches~1.1.Thus all TMaod of amplitudes < 2。

In order to ensure the limitation, which properly separates out by the point of these dither tables (increases with DitherRange To increase Nd), and segmentation speed is limited to constrain | DitherOA |.

Note that the scaling in TMaod is likely larger than 1, it is not intended that the change in overlapping, because The scaling item in TMaod compensates other scaled results.For example, the longer bevel edge of the shake vector (because of velocity compensation, is shown in figure 38) occur to reduce an overlapping, but when these points are used for the workpiece surface by reality, which lines up them directly Line and the geometry return normal.Identical algorithms are used for local lens aberration and scale (TMfield > 1 normalization local lens Scaling < 1).Thus, which can point out nominal point interval without considering TMaod scaled results.

6. vector processing is summarized

In some instances, the processing that can be summarized in as follows in vector pattern.For each of application processing segmentation According to the following steps.Figure 48 (being discussed below), which is described, is converted into the FPGA using for vector is processed segmentation data The processing flow of 1028 order.Technical staff will recognize at disclosure from there, can implement this with different order Certain steps in the following steps.

Step 1: the application is divided into independent processing to be segmented by the SCC 1010 using following rules: independent segmented for circle Needed for arc；All segmentations are less than maxSegmentLength (~1mm, for calibrating)；And to be greater than 12.5 μ sec to all points (for the particular instance, processing supports every segmentation by the renewal rate of 12.5 μ sec to Duan Jiagong, will support with 2m/sec's 25 μm of minimum section length；Certainly, different renewal rates can be used in other examples).

Step 2: the SCC 1010 determines that this is required using the manuscript and the trench geometry (width, depth) Machined parameters (GridWidth and Fluence).

Step 3: the SCC 1010 determined using the shake figure corresponding to required mesh width dither table parameter (Nd, Kw).Notice that Nd continues to have for each segmentation and Kw may be carried out interpolation in conversion segmentation.

Step 4: the SCC 1010 has used the shake figure to determine the dither table parameter tableAddress for corresponding to Nd And tableLength.

Step 5: the SCC 1010 has used the shake figure to calculate EffectiveWidth.Due to it can be carried out to Kw in It inserts, interpolation can also be carried out to EffectiveWidth in conversion segmentation.

Step 6: the SCC 1010 calculates Dosage=Fluence*EffectiveWidth.It can be in conversion segmentation to it Carry out interpolation.

Step 7: the SCC 1010 is based on the segmentation circular-arc radius (if necessary) and calculates Kb.

Step 8: the SCC 1010 specifies segmentation speed (to keep on that segment based on the one or more of following limitation It is constant): the radium-shine power limit speed of a., based on the power diagram and the required dosage；

B. according to the required mesh width, the AOD rate limitation of AOD sizes and third level filter transient state amplitude；c. According to the AOD rate limitation of the largest interval (=Taod* (Nd-1) * Vel) between shake row, possibility < 0.35*Deff； May have with e. segmentation compared with low velocity, if necessary to be used to meet the minimum split time demand proposed in step 1.

Step 9: the SCC 1010 is by the processing and position distribution data transmission to the DSP 1026.(in this example not Use the form parameter Ks).The data for being transmitted to the DSP 1026 from the SCC 1010 includes segment endpoint XY coordinate, sectional circular Arc center XY coordinate and radius, multiple track samples (~speed), dither tableAddress, dither TableLength, Dosage, Kw and Kb.

Step 10: the DSP 1026 calculates position of center line data (Xc, Yc), with the update cycle of Tcmd, and uses The third level filter processes the data.

Step 11: the DSP 1026 calculates new shake (Kw) and processing (Kp, Kb) parameter, each shake row is primary (Nd*Taod seconds every), if necessary to carry out interpolation (in conversion segmentation).

Step 12: the DSP 1026 is based on trajectory angle and speed calculates the TMaod transition components.

Step 13: the inferior grade order is issued the FPGA 1028 to control AOD operation by the DSP 1026.Due to the DSP For 1026 and FPGA 1028 with different update period (Tdsp vs.Taod) operation, each DSP period transmits sufficient data to permit Perhaps the FPGA 1028 reaches Tdsp seconds next (that is, entire Tdsp/Tcmd data set) to process.

B. grating is processed

Grating processing is similar to vector and processes, and in addition to 2D grating table is controlled for AOD, rather than is used in vector processing 1D table.

In one example, coaxially implement grating processing together with vector processing, which obviate access stop position again Time waste hovered knot the elapsed time minimized between vector write operation and crosspoint processing with minimizing radium-shine light beam Fruit.In addition, the inspection of the big field scan lens aberration characteristic, which points out that big grating pattern (e.g., 200 μm wide) may have, to be made PGC scaling and rotation amendment, especially if when being intersected by groove to them.

According to the LDA system of an example, by using the identical hardware structure for grating and vector processing Mixed vector and grating processing.Due to specifying unique dither table for each processing segmentation, raster manipulation is seamlessly tied It closes into the processing flow as another table.

Generally zero is set during grating processing by the speed of the nominal light-beam position.In addition, if machined Biggish grating region (major part for consuming the AOD range), allowing the third level filter to be arranged to minimum should Increased AOD third level deflection.Generally, in some instances, this is the sequence according to 0.25 millisecond.

Scaling/rotation amendment can be provided as one group PGC, as described in previous dither rotation discussion.It is right It is processed in grating, which includes the rotation of the grating pattern in galvanometer XY coordinate framework, and by the SCC 1010 are transmitted (rather than be based on path velocity by the DSP1026 and calculated) for each grating pattern.The rotation can Crosspoint grating pattern can be oriented relative to the surrounding track for being used to correctly intersect dot format.

In order to consider to scan field distortion, which also identifies PGC amendment, XY scanning field position based on grating and The galvanometer calibrates data.In some instances, this has modified the mistake for reaching 5-6% in X/Y error on the grating region Difference.This may be not correct error during grating processing.Pay attention in vector processing, is clearly used for scanning field calibration Segment endpoint；Nonlinearity erron between the terminal on short distance (~1mm) is smaller.

1. grating processing is summarized

The processing in grating mode can be summarized as follows.The following steps can be taken for each of application processing segmentation, schemed 49 (being discussed below) summarise the processing stream of the order for grating processing segmentation data to be converted into the FPGA 1028 Journey.Technical staff should recognize at disclosure from there, can implement certain steps in the following steps with different order Suddenly.

Step 1: the SCC 1010 is calculated grating pattern during " conversion " of the application and is come using some independent patterns Construct grating table.In some cases, single grating " position " can be by several smaller grating patterns (e.g., circle pad+at each angle " label " at degree) it constitutes.These can be incorporated to biggish pattern, or processed respectively.Pay attention to existing in application process At different rotation angles, some grating patterns can be used for several times.

Step 2: the SCC 1010 identification be used for all grating patterns rotation angle, such as CAD (computer aided design, For the specific application) coordinate.

Step 3: the SCC 1010 is by grating table data download to the FPGA 1028, before the application starting.

Step 4: the SCC 1010 is that each stop position creates TMdither conversion.The PGC, which corresponds to, to be sent to The TMaod of the FPGA 1028 is converted.TMdither conversion includes the stop position in CAD coordinate, the workpiece alignment Rotation, the locally scanning field distortion PGC, and the additional scaling adjustment selectively specified by the user.

Step 5: the SCC 1010 forms grating processing segmentation, and the duration of the grating is used for using zero velocity.Such as The fruit user specifies the repetitive rate (e.g., for better machining control) of the grating, can be linked together multiple segmentations.

Step 6: the SCC 1010 may include the zero velocity segmentation of guidance and hangover, has and arranges to be used to during the time to allow The third level filter arrangement prevents the excessive AOD of big grating pattern from deviating.

Step 7: the SCC 1010 calculates the required fluence based on the specified grating depth and the manuscript.

Step 8: the SCC 1010 is nominal to calculate based on the grating spot pitch and the optional Kw scaling adjustment factor Power.

Step 9: grating parameter is sent the DSP 1026 by the SCC 1010.Notice that the DSP 1026 may be by " dosage " Be construed to " power " for raster-segment, by its geometry scaling be incorporated to the PGC rather than Kw, and without using the slope. " grating mode " identification can be used for being marked as different data structures.The grating of the DSP is sent to from the SCC 1010 Parameter may include segment endpoint XY coordinate (being equal for zero velocity), and segmentation circular-arc center XY coordinate and radius (do not make With), the quantity (~raster time) of track sample, raster tableAddress, raster tableLength, power (replaces Change in dosage) and TMdither (field distortion & grating angle).

Step 10: the DSP 1026 sends the FPGA 1028 for above-mentioned grating parameter and is used to process.Since the grating is seen Get up as shake, will not need that any special pattern is arranged.

2. the grating processing of tiling

The variation of grating processing can be used in certain situations.Figure 39 is described with figure and is processed according to the tiling grating of an example Example.In Figure 39, two pads 3910,3912 are close together and are connected by short track 3914.The whole region It is too big for single grating pattern, and the pad 3910,3912 can be used for typical pad-rail close to distribution together Mark-pad job sequence.Can the scraping region, although the radium-shine duty factor will be lower due to the shape of the machining area.

Alternative according to an example is for being two grating patterns 3916,3918, quilt by the region processing Tiling and overlapping are to provide the transition region 3920 between two grating regions 3916,3918.It can be on the SCC 1010 Define the grating pattern 3916,3918 according to: will arbitrary " drop block " (that is, region with arbitrary shape and/or size) point Solution is the grating region 3916,3918 of several overlappings, is had equal to the required transition length 3922 (e.g., about 40-50 μm) It overlaps, and around the boundary of each drop block beyond the transition length 3922；It is filled up often using the dot grid with consistent amplitude One region；FIR filter is applied to oblique deascension grating amplitude (e.g., the class in the transition region 3920 along each dimension Like the scraping filter)；In each grating region, without the data except the drop block region, which results in good definition Drop block region, there is slope transition region between piece；And each raster pattern is downloaded and executed at the suitable position Case.

This may be useful for the smaller droplet block that can be included in several grid fields, thereby increases and it is possible to be faster than low close in Figure 39 Spend pattern scraping because only the required point can by grating (skipping all white spaces), and because between AOD only It is mobile only to make several galvanometers.

The pattern (using suitable rotation) of repeatability can be repeated, it is opposite with recalculating for each new SkiveData. Certainly, identification " equivalent pattern " may be difficult.

C. shaving action

Scraping is that material is removed to the processing of big region (rather than thin groove).Mode summarized below, which allows to shake, opens up The wide line by than use these nominal dots more preferably in a manner of remove material, while avoiding to be originated from using wide in this way line " pel " result.For example, Figure 40 describes the dot matrixed error according to certain examples due to the wide line scraping with figure.In Figure 40 In, it include being fallen using be fine to that the light beam for being capable of providing sufficient resolution ratio processed enough by the example of processing drop block (left side) The edge of the circle.Excessive pel is resulted in using the angle pencil of ray for being used to process this feature (the right) and unacceptable resolution ratio is damaged It loses.

In some instances, it might be useful to form side and the terminal ramp of the scraping light beam.These " relay ramps " are permitted Perhaps the adjacent scraping access correctly overlaps and generates unified fluence (e.g., for unified deep-controlled) on the workpiece simultaneously For tolerance is supplied to light-beam position error.In addition, slope in the terminal of the scraping light beam will be by the scraping region Groove is needed in the case where intersecting and (consistency and tolerance again, is supplied to location error).

In one example, dosage and shape control to implement such forming of wide scraping light beam.However, the wide line Pel problem makes which difficult, if remained unchanged the line width during processing.

In another example, which is changed by processing " grid of pel ", and filter is used for this Grid is to generate the required side and terminal ramp.

Figure 41 describes the scraping region 4100 for being divided into lengths 4102,4104,4106 according to an example with figure.When When only showing three band 4102,4104,4106, it can be used additional belt to complete the entire scraping region 4100.It will be common Or arbitrary " drop block " is converted into pel grid 4110, has the figure based on the basic point size (overlapping for e.g., needing > 60%) Minimum interval between member 4110.Scraping object may have an extended axis, can be aligned in the scraping beam trajectory (e.g., as indicated by the arrow 4112 in the example of Figure 41, from left to right for the first lattice band 4102).

It is likely less than in the quadrature axis (" intersecting axle ") along the grid interval of the general beam direction track (being somebody's turn to do " coaxial ") Required interval, be based on speed, jitter points renewal rate and jitter points quantity.Establish the region that the dot grid makes the scraping Accurately summarized, especially if when other any features will intersect the scraping region.The thus last dot grid interval Based on the outer dimension.Equally, according to certain examples, since the shake row execution time is fixed and quantifies (=Taod*Nd), Adjusting the scraping path velocity makes the scraping shake row of the integer correctly insert the scraping region.

Then by the grid dividing, at independent access, (Figure 41 is shown as 4102 (bands 1), 4104 (bands 2) and 4106 (bands 3) it, is shown as being divided by dotted line for descriptive purpose).During each access, needed for shaking the light beam to create Width.Generally, this there may be unacceptable pel errors, as shown in figure 40.However, if it is possible to as same axle position The function set reduces the pattern of the shake, can be avoided such pel.

In order to avoid pel, which is continued to be loaded into the FPGA 1028AOD controller, be had by binary pattern Each specified " shake row ".Then the pattern is filtered to generate side and terminal ramp.Filter is shown in FIG. 42 The result of grid after wave.Figure 42 is described with figure according to the unfiltered grid (left side) of the scraping jitter points of an example and corresponding The filtered grid (the right) of scraping jitter points.Figure 43 is described with figure corresponds to the band in Figure 41 according to an example The fluence of 4102,4104,4,106 first three scraping light beam is distributed.First Figure 43 10 is shown corresponding to first band 4102 fluence distribution.Second Figure 43 12 shows the fluence distribution corresponding to the supplement of second band 4104.Third figure Show the fluence distribution of the supplement corresponding to third band 4106.4th Figure 43 16 is shown corresponding to the band 4102, The fluence of 4104,4106 combination fluence is distributed.

In some instances, jitter direction (the intersection axis direction relative to the scraping beam trajectory) is based on the scraping light beam The direction of track, which is taken to from one between next access, to be switched.For example, referring again to Figure 41, for the first band 4102 Scraping beam trajectory 4112 is that from left to right, the scraping beam trajectory 4114 for second band 4104 is from right to left, to be used for The scraping beam trajectory 4116 of the third band 4106 is again from left to right.As the general beam direction track is in each scraping access Between reverses direction, Figure 41 describes the jitter direction along each shake row (as pointed out by row 4118,4120,4122 As) also change direction.First access is trembled as the scraping beam trajectory 4112 is from left to right moved with first Dynamic direction 4118 (e.g., from bottom to top) processing is shaken in each of the first band 4102 goes.For the alternate path, As the scraping beam trajectory 4114 moves from right to left, existed with second jitter direction 4120 (e.g., from the top to the bottom) processing Each of second band 4104 shake row.For the third path, with the scraping beam trajectory 4116 again from a left side to It moves right, is shaken with first jitter direction 4122 (e.g., from bottom to top) processing in each of the third band 4106 Row.According to the switching of the jitter direction in the direction based on the scraping access of certain examples, influence the shaving action control and Quality (depth for e.g., influencing the scraping).In some instances, allow the user to choose whether to overturn or protect during the scraping Hold the jitter direction (direction e.g., based on the scraping access).

The scraping process outlined below summarised according to an example.

Efficiency of the selection of the scraping line width for optimization.Two example scenarios discussed below.

Without any arrangement after first suppositive scenario is mobile between scraping row, as described in Figure 44 and Figure 45.Figure 44 describe the scraping example according to an example with figure.In Figure 44, mutually by multiple scraping rows 4410,4412,4414,4416 It is separated for describing purpose.However, as discussed above relative to Figure 41 and 43, technical staff should it is disclosed from there in Hold it is understood that these scraping rows 4410,4412,4414,4416 can also be mutually adjacent (or mutually partly overlapping).Figure 45 figure Describe the light beam order during the scraping according to an example.As processed each scraping row 4410,4412,4414,4,416 one Sample, the position third level AOD is (after the third level filter has arranged) placed in the middle, and the AOD subsystem 1042 shake exists Around the center, pass through the half of ± line width.As the beam positioner completes each row and moves on to next line, It is (capable and next in this prior equal to the additional deflection in the AOD of line width half to create (because third level section) AOD deflection is separated between row).Thus the maximum AOD deflection after the jump to next line consumes an entire line width Degree.It " is not arranged " in mode at this, although not needing arbitrarily to arrange after the jump to next line, selects the length of every a line Degree is so that be sufficiently long to that the third level filter can allow for be arranged after previous row jump.Scraping based on the selection Width is cut, shorter row length can be still using additional " arranging to be segmented " to provide the time.Note that even if well should Scraping width is arranged under maximum AOD range, and the settable power (and thus speed) is to adapt to entirely deflect model at it The AOD efficiency for the worst condition placed, thus reduce the efficiency of which.

Second scenario allows using the entire AOD width, but has and be always inserted into allow the third level transient state to exist The arrangement segmentation that the processing of scraping row is failed before starting.This allows using wider scraping width.When this may be moved by every row Except more many areas is come when increasing scraping efficiency, by increase scraping width acquisition efficiency by wider range of deflection compared with Low AOD efficiency offsets (assuming that the system, which allows power as AOD, deflects the function of width to increase).

The model of scraping efficiency from the two scenes is described in figures 4-6 can, is that material shifting is described according to an example The figure of removal rates (region per second) vs. scraping row length (scraping length) and width (line spacing).Figure in Figure 46 assumes 10 μm of scrape depth and 25 μm of point size.This is indicated, for being somebody's turn to do " not arranging " scene effect for the AOD in the example It is close in rate to be somebody's turn to do " arrangement " scene.This can design with different AOD and be changed, which has more efficient or broader deflection model It encloses, or for different scrape depths or nominal point size.Notice that this " does not arrange " situation to operate in the width range more limited, Because the additional AOD in the jump process between being expert at is deflected.In this example, which is about 50 μm.

Figure 47 describes the side view of the scraping row geometry according to an example with figure.Figure 47 is described before filtering Scraping line width 4710 and shake width 4712 after the filtering.Filtering produce inclined side wall 4714 (e.g., for The intersection of adjacent scraping row or other features), in the transition region with transition width 4716.The gradient and the conversion are wide Degree 4716 is based on the filtering (e.g., CrossAxisTaps* spot pitch).

1. shaving action is summarized

The processing in scrape mode can be summarized as follows.Technical staff will recognize at disclosure from there, can be different Certain steps sequentially to implement the following steps.Figure 50 (following discussion), which is summarised, is transformed into this for grating processing segmentation data The processing flow of the order of FPGA 1028.

I. preliminary step

Step 1: the SCC 1010 identifies scraping region, the individual course that can be replaced by the application file.

Step 2: the scraping regioinvertions at the grid of binary bit point, are had the intersection based on size by the SCC 1010 Between centers is every and the coaxial spaced based on speed and jitter parameter.

Step 3: the SCC 1010 is by the scraping Mesh Conversion at one group of row.

It is shown in Figure 46 that line width can be optimized to (width vs. efficiency tradeoff).Note that scraping width is because create this The filtering of bevel edge and increase, therefore the width optimization calculate may be taken into account.Edge transition may be about 40 μm wide It spends (for providing 5% depth error of 2 μm of location errors)；The extension etc. of filter edge at the unfiltered nominal edges In the 1/2 of the transition length.

B. scraping shake row interval (in the coaxial direction) meets largest interval (point overlap > 60%), in power limit Process velocity and each scraping segmentation in the capable demand of integer shake.In addition, the shake row interval may be compatible with The required coaxial transition length and the specified coaxial boxcar filter (as seen below).Since boxcar filter is grasped Make on integer shake row, which may be provided in the integer row in the transition region.It if necessary, can be slight Amplify the transition length to allow the situation, because of any corresponding crosspoint of matched transition length.

Step 4: the SCC 1010 point out the length of boxcar average (CrossAxisTaps and OnAxisTaps) with Create the required slope on edge and terminal.

A.CrossAxisTaps and OnAxisTaps is odd number (midpoint+surrounding point to).

B.CrossAxisTaps and OnAxisTaps is possible different, based on the intersecting axle interval (shake control) and coaxially It is spaced (being controlled by Nd*Taod*vel).The transition region may be with for the identical of crosspoint, and e.g., about 40 μm (for 2 μm 5% depth error of location error).

C. because OnAxisTaps is the value being rounded, the timing (passing through speed and/or Taod) of the shake row may need It is adjusted to meet the demand in the final position and relay ramp width.

Step 5: for each scraping row, which creates one group of SkiveData word.

A. each SkiveData word illustrates one group of amplitude point of every shake row, has the undressed point for being set as zero.

B. the length of the SkiveData word is equal to the dither table length (reaching at 32 points) for scraping.

C. in each terminal of SkiveData word, settable " CrossAxisTaps " point is to consider that the boxcar is filtered The width of wave device, as shown in Figure 47.

D. (opposite to be somebody's turn to do by the jitter points of " CrossAxisTaps " quantity by the radium-shine process excursion in the intersecting axle The terminal of dither table).The scraping dither table considers this offset.The scraping dither table also includes enough as described above Zero padding entrance.

Step 6: the SCC 1010 calculates processing segmentation position.Along this coaxially by the shake row of " OnAxisTaps " quantity To deviate the radium-shine processing of reality of (delay) each scraping row.Processing segmentation geometry considers this offset.

Step 7: the SCC 1010 has used the manuscript and the trench geometry (width, depth) to determine that this is required Machined parameters (GridWidth and Fluence).

Step 8: the SCC 1010 has used the shake figure to determine the dither table parameter for corresponding to the required GridWidth (Nd, Kw).

Step 9: the SCC 1010 has used the shake figure to determine the dither table parameter tableAddress for corresponding to Nd And tableLength.

Step 10: the SCC 1010 has used the shake figure to calculate EffectiveWidth.Due to may interpolation Kw, Upper interpolation EffectiveWidth can be segmented in conversion.

Step 11: the SCC 1010 calculates Dosage=Fluence*EffectiveWidth.It can be in conversion segmentation Shang pair Its interpolation.

Step 12: the SCC 1010 has used the power diagram to be determined as the permitted maximum speed of the segmentation, and being based on should Required dosage.Segmentation may have lower speed, when meeting the minimum split time demand if necessary.Arbitrarily it is being segmented On speed is remained unchanged.

Step 13: the SCC 1010 scales the dosage and passes through

Kskive=1/ (CrossAxisTaps*OnAxisTaps).

Step 14: the SCC 1010 notifies the DSP 1026 to enter scrape mode.The FPGA 1028 is arranged in the DSP 1026 Into scrape mode.

Ii. operation-time processing

Each segmentation during shaving action is according to the following steps.

Step 1: the SCC 1010 is that each scraping row creates processing segmentation, and the following information is transmitted to this DSP 1026: segment endpoint XY coordinate；Segmentation circular-arc center XY coordinate and radius；The quantity (~speed) of track sample；Scraping tableAddress；Scraping tableLength；Dosage；Kw；SkiveData；OnAxisTaps；CrossAxisTaps；With Scrape mode notice.

Step 2: the DSP 1026 processes the stream data at the SCC 1010 with vector pattern, sends the data to The FPGA 1028.

Step 3: the FPGA 1028 creates shake row vector, has the non-zero material point for being set as the prescribed dose.

Step 4: the FPGA 1028 executes intersecting axle and coaxial boxcar average filter on the shake row, as follows It is shown:

A. boxcar filter is run when (each shake row) in the intersecting axle data is in its arrival.Each point Each boxcar filter be the point and should ± (CrossAxisTaps-1)/2 surrounding point (have in the vector terminal The zero padding at place) summation.Any scaling is not needed (dosage has been scaled by Kskive in advance).

B. to these, filtered jitter row has been lined up.

C. the shake row is processed from the queue.It is filtered using coaxial boxcar, by will be around each row and this (OnAxisTaps-1)/2 row (zero padding beginning as needed with terminal row) is added.Again, because the Kskive is scaled in advance Without any scaling.

Step 5: paying attention to still using angle and dosimetry parameter (as in regular dither), but be not shape or slope (boxcar filter uses shape；Scraping is constrained to straight line, therefore without slope).

Step 6: adjustment is synchronous to consider the delay at the boxcar filter.

Figure 48 is the block diagram generated for the AOD order of vector processing described according to an example.As discussed above Equally, which will be using independent processing segmentation be split into, each has corresponding segmentation data 4810.Use material Expect table 4812 (e.g., seeing Fig. 9), geometry (e.g., the depth of groove of the SCC 1010 processing at the segmentation data 4810 And width) and data from manuscript 4814 to determine the machined parameters such as GridWidth and Fluence.The SCC 1010 Dither table parameter Nd and Kw are determined using shake Figure 48 16 to process the mesh width.It in some instances, is each point Nd is remained unchanged but can be segmented upper interpolation Kw in conversion by section.The SCC 1010 is then corresponded to using shake Figure 48 16 with determining In dither table the parameter tableAddress and tableLength of Nd.The SCC1010 also uses the shake figure to calculate EffectiveWidth(Weff).It, can also be in conversion segmentation by EffectiveWidth interpolation because may be to Kw interpolation.Such as Shown in Figure 48, which calculates Dosage=Fluence*EffectiveWidth.It can also be in conversion segmentation to agent Measure interpolation.DitherRange is also supplied to the DSP 1026 by shake Figure 48 16.

The dosage is supplied to power diagram 4818 to the SCC 1010 and specified segmentation speed MaxVel (is kept on that segment It is constant).Segmentation speed MaxVel is potentially based on radium-shine power limit speed and/or one or more AOD rate limitation.It is radium-shine Dosage of the power limit speed based on the power diagram 4818 and the calculating.It is wide that the AOD rate limitation is potentially based on the required grid Degree, the AOD size and the third level filter transient state amplitude.In addition, or in other instances, which may Based on the largest interval (=Taod* (Nd-1) * Vel) between shake row, pressed according to certain embodiment party as < 0.35*Deff. In some instances, segmentation may have compared with low velocity, when if necessary to be used to meet minimum split time demand.

The track SCC generates mould group 4820 and receives XY light from the segmentation data 4810 and the specified segmentation speed MaxVel Beam coordinate.If slope is segmented for circular arc, which generates mould group 4820 and calculates slope based on the segmentation circular-arc radius Parameter Kb.The track SCC generates mould group 4820 and also generates segmentation track data (e.g., segment endpoint XY coordinate, segmentation circular-arc center XY coordinate and radius and multiple track samples).

The SCC 1010 has used the segmentation data 4810 to generate and correct scanning field distortion and nominal galvanometer alignment TMfield conversion 4822.The SCC 1010 generates TMframe conversion 4823 also the AOD coordinate framework is aligned in the inspection Flowmeter coordinate framework.As shown in Figure 48, which converts TMfield conversion 4822 and the TMframe 4823 are multiplied to determine the TMdither transition matrix.

In the DSP 1026, the track the DSP generation use of the mould group 4824 segmentation track data, DitherRange, and from The Kw received at the SCC 1010 calculates position of center line data (Xc, Yc), at the more new period of Tcmd.Although not existing It is shown in Figure 48, the DSP 1026 is also using third level filter (referring to such as the third level in Figure 22 and 51 in some instances Filter 2205) process the data.The track DSP generates 4824 output speed of mould group, and AOD deflects coordinate, and shake vector Ratio/rotation.By the AOD deflection coordinate be supplied to the TMtert conversion 4832, output frequency offset command Fdev0 and Fdev1。

The every shake row of the DSP 1026 is (Nd*Taod seconds every) to calculate primary new shake (Kw) and processing (Kp, Kb) parameter. If the segmentation is conversion segmentation, which calculates the new shake Kw parameter using interpolation 4826.For conversion point Section, the DSP 1026 is also using interpolation 4828 to determine dosage.As shown in Figure 48,1026 use of DSP is by the track DSP The dosage calculated of mould group 4824 and speed are generated to determine the workpiece surface power for being supplied to Kpwr mould group 4830.The Kpwr mould The workpiece surface power is converted to standardization power command Kp by group 4830.

As shown in Figure 48, which is based on TMdither and shake vector ratio/rotation calculates TMaod conversion Component, as described above.

The DSP 1026 sends the FPGA 1028 for the inferior grade order to control AOD operation.

Figure 49 is the block diagram for describing the generation of the AOD order for grating processing according to an example.As discussed above Equally, which will be using independent processing segmentation be split into, each segmentation has corresponding segmentation data 4810.It should SCC 1010 calculates grating pattern and establishes grating table using some independent patterns during " conversion " of the application.Segmentation money Material 4810 includes the grating rotating angle for the grating pattern.Each raster-segment has zero velocity.

Using material table 4812, which processes grating geometry (e.g., the depth from the segmentation data 4810 And width) and data from the manuscript 4814 to determine adding such as RasterID (e.g., corresponding to mesh width) and fluence Work parameter.The SCC 1010 processes the RasterID using shake Figure 48 16 to determine dither table parameter Nd and Kw, with fluence Ratio is together.As shown in Figure 49, user optionally adjusts Kw.The SCC 1010 is also using shake Figure 48 16 with determination Grating table parameter tableAddress and tableLength corresponding to Nd.As shown in Figure 49, which passes through in the future From the fluence of the material table 4812 multiplied by the fluence ratio as provided by shake Figure 48 16 to calculate required fluence.

The track SCC generates XY light beam coordinate of the reception of mould group 4820 at the segmentation data 4810.The track SCC is raw Segmentation track data is generated at mould group 4820, wherein it is supplied to the track DSP and generates mould group 4824.

As shown in Figure 49, which comes from the segmentation using TMfield conversion 4822, TMframe conversion 4823 Grating rotating data at data 4810, and Kw at shake Figure 48 16 is to calculate the TMdither transition matrix.It should TMdither conversion corresponds to the TMaod conversion being sent at the FPGA 1028.

As shown in Figure 49, which can be construed to dosage to be used for " power " of raster-segment.The Kpwr mould group The power is converted to standardization power command Kp by 4830.

The track DSP generates mould group 4824 and exports AOD deflection coordinate.AOD deflection coordinate is supplied to the TMtert to turn 4832 are changed, output frequency deviates order Fdev0 and Fdev1.

Figure 50 is the block diagram for describing to be generated according to the AOD order for shaving action of an example.As discussed above one Sample, the SCC 1010 will be using independent processing segmentation be split into, each segmentation has corresponding segmentation data 4810.At this In example, which identifies scraping region, can be replaced in the independent stratum in application file.The SCC 1010 will Each scraping regioinvertions has the intersecting axle interval based on size, and based on speed and tremble at the grid of binary bit point The coaxial spaced of dynamic parameter.As discussed above, the SCC 1010 is by the scraping Mesh Conversion at one group of row.

Using the material table 4812, trench geometry of the SCC 1010 processing from the segmentation data 4810 is (e.g., deep Degree and width) and data from manuscript 4814 to determine the machined parameters such as GridWidth and Fluence.The SCC 1010 are used to determine dither table parameter Nd and Kw using shake Figure 48 16 to process the GridWidth.In some instances, right Nd is remained unchanged in each segmentation but upper interpolation Kw can be segmented in conversion.The SCC 1010 then uses shake Figure 48 16 To determine the dither table parameter tableAddress and tableLength that correspond to Nd.The SCC 1010 is also schemed using the shake To calculate the EffectiveWidth (Weff).Because can interpolation Kw, upper interpolation EffectiveWidth can also be segmented in conversion. As shown in figure 50, which calculates Dosage=Fluence*EffectiveWidth.Upper interpolation can also be segmented in conversion Dosage.DitherRange is also supplied to the DSP 1026 by shake Figure 48 16.

The dosage is supplied to power diagram 4818 to determine the maximum speed MaxVel of the segmentation (in this point by the SCC 1010 It is remained unchanged in section).In some instances, if necessary to meet minimum split time requirement, segmentation may have lower speed Degree.

The track SCC generates mould group 4820 and receives XY light beam coordinate and the maximum speed from the segmentation data 4810 MaxVel.The track SCC generates mould group 4820 and points out the boxcar average (CrossAxisTaps and OnAxisTaps) Length is to create the required slope on edge and terminal.Thus, during foundation, which generates mould group 4820 will Skive Taps is supplied to the FPGA1028.Each corresponding scraping row, the track SCC generate mould group 4820 and also create one group SkiveData word.The track SCC generates mould group 4820 and generates segmentation track data, may include processing segmentation position.It will be every The radium-shine processing of the reality of one scraping row coaxially deviates (delay) by the shake row of " OnAxisTaps " quantity along this.It should add Work point section geometry considers the offset.

As discussed above, which scales the dosage, passes through Kskive=1/ (CrossAxisTaps* OnAxisTaps)。

As shown in Figure 50, which is multiplied to TMfield conversion 4822 and TMframe conversion 4823 really The fixed TMdither transition matrix.

In the DSP 1026, the track DSP generate mould group 4824 use the segmentation track data, DitherRange and For the Kw received from the SCC 1010 with output speed, AOD deflects coordinate and shake vector ratio/rotation.The AOD is deflected Coordinate is supplied to TMtert conversion 4832, output frequency offset command Fdev0 and Fdev1.

As shown in Figure 50, which generates the dosage calculated of mould group 4824 and speed by the track DSP To determine the workpiece surface power for being supplied to Kpwr mould group 4830.The workpiece surface power is converted to this by the Kpwr mould group 4830 Standardize power command Kp.The DSP 1026 calculates this based on TMdither and shake vector ratio/rotation as discussed above TMaod transform components.

The inferior grade order is sent to the FPGA 1028 to control AOD operation by the DSP 1026.The FPGA 1028 wound Shake row vector has been built, the non-zero material point for being set as the prescribed dose is used.The FPGA1028 is also executed on the shake row Intersecting axle and coaxial boxcar average filter.

Figure 51 is the block diagram for describing to calibrate data process 5100 according to the galvanometer and AOD of an example.In the galvanometer In AOD calibration data process 5100, which provides (the example light beam point in such as Figure 22 of SCC trace specification 5110 Cloth 2210), XY segmentation coordinate (e.g., in workpiece surface coordinate) is compared to the scan transformation 2203.Scanning field conversion Modified XY segmentation coordinate (e.g., in original galvanometer coordinate) is then supplied to the track DSP generation unit 4824 by 2203, It calculates and is sent (e.g., in original galvanometer coordinate) to the specific beam trajectory of the third level filter 2205 and shake ginseng Number.As discussed above, the third level filter 2205 is by original XY galvanometer control command (e.g., in original galvanometer In coordinate) it is supplied to galvanometer controller 5114.Also by XY AOD order, (be given to may be in original to the third level filter 2205 Increased optional XY galvanometer errors correction term in beginning galvanometer coordinate) it is supplied to TMtert conversion 4832.Then in the future It is supplied to from the output (e.g., similar ch0 and ch1AOD order) of TMtert conversion 4832 and produces the final AODRF order FPGA 1028.

The SCC 1010 also merges rotation conversion TMframe 5118 with local scanning field distortion amendment TMfield4822 To generate TMdither transition matrix.The TMdither transition matrix is then incorporated in shake rotation/scale conversion 5122 (e.g., Rdither and KwCorr) is provided to the FPGA 1028 to generate AOD order transition matrix TMaod conversion.It should SCC 1010, can also using SCC shake data 5124 (e.g., in workpiece surface coordinate) to generate MHzPer μM of scale factor It is included as being pre-loaded into data in the shake (or grating) table 1110.The FPGA 1028 will come from the dither table 1110 data merges TMoad conversion to generate shake and grating data, and wherein the FPGA 1028, which merges, comes from the TMtert The AOD order at 4832 is converted to generate the AOD RF order.

Skilled worker be familiar with it should be understood that many variations can be carried out to the details of examples detailed above without departing from the present invention Cardinal principle.Therefore the scope of the present invention should be determined only by following claims.

Claims (19)

1. a kind of radium-shine system of processing, the radium-shine system of processing include:

First positioning system, the first phase of the beam path for assigning the radium-shine light beam track along the surface relative to workpiece To the first positioning system of movement；

One or more processors, for determining second relative motion in the radium-shine light beam path along multiple shake rows, Second relative motion with the predetermined angular superposition relative to the beam trajectory on first relative motion, it is described The determination of second relative motion include for for it is the multiple shake row each retain the predetermined angular along the light The compensation changed in the process velocity of beam trajectory；

Second positioning system, for assigning second relative motion in the radium-shine light beam path；And

Radium-shine source, for being shaken multiple radium-shine light beam impulse ejections at multiple capable positions described in along the multiple Workpiece is used to widen groove with the defined direction of the predetermined angular.

2. the system as claimed in claim 1, which is characterized in that the predetermined angular is substantially perpendicular to the light beam rail Mark.

3. the system as claimed in claim 1, which is characterized in that first positioning system includes from including galvonometer driven mirror With one or more beam positioner selected at the group of quick manipulation mirror.

4. the system as claimed in claim 1, which is characterized in that two positioning system includes from including acoustics-deflection optical Selected one or more beam positioner at the group of device and electro-optical deflector.

5. one kind is used to carry out micro-machined radium-shine system of processing to workpiece, the system comprises:

Radium-shine source, to generate radium-shine light beam for machining feature in the workpiece；

Acoustics-optical deflector subsystem, is configured to radium-shine light beam described in range of deflection angles intrinsic deflection；

Locator is configured to further deflect the laser light deflected through the acoustics-optical deflector subsystem Beam is to assign the movement of the radium-shine light beam point position along the machining locus relative to the workpiece；

Scanning lens, to focus on the radium-shine light beam at the workpiece；With

Relay lens is located to receive relative at the selected position of the scanning lens from the acoustics-optics The radium-shine light beam of deflector subsystem and the light beam fulcrum of the radium-shine light beam is relayed to the locator, wherein in the work It is inclined by the acoustics-optical deflector subsystem institute that the incidence angle of radium-shine light beam at the surface of part corresponds to the radium-shine light beam The angle turned.

6. radium-shine system of processing as claimed in claim 5, which is characterized in that the locator is from including galvonometer driven It is selected in the group of system and quick manipulation mirror system.

7. radium-shine system of processing as claimed in claim 5, which is characterized in that the laser light at the surface of the workpiece The incidence angle of beam corresponds to the radium-shine light beam and is independently of institute by the angle that the acoustics-optical deflector subsystem is deflected State the movement for the radium-shine light beam point position that locator is assigned.

8. a kind of workpiece for the metal layer containing dielectric layer and the dielectric layer that overlaps carries out micro-machined method, described Method includes:

Generate radium-shine light beam；And

It guides to process the workpiece in the radium-shine light beam to the workpiece, wherein the guiding includes:

In the first procedure of processing, guides and the metal layer is penetrated with cutting in the radium-shine light beam to the workpiece；And

In the second procedure of processing, guide to process the dielectric layer in the radium-shine light beam to the workpiece,

Wherein, in second procedure of processing, the radium-shine light beam is in order through the first acoustics-optical deflector and the The transmitting of two acoustics-optical deflector,

Wherein when the radium-shine light beam is transmitted through the first acoustics-optical deflector during second procedure of processing When, it is sufficient to the first linear frequency-modulated sound wave of radium-shine light beam diffraction in a first direction is generated only at first sound In-optical deflector, and

Wherein when the radium-shine light beam is transmitted through the second acoustics-optical deflector during second procedure of processing When, it is sufficient to the second linear frequency modulation acoustic wave of radium-shine light beam diffraction in a second direction is generated only at the rising tone In-optical deflector, wherein the second direction is not parallel with the first direction or reversed.

9. method according to claim 8, which is characterized in that first procedure of processing and second procedure of processing are on edge Relative to the workpiece beam trajectory the radium-shine light beam single access in be performed.

10. method according to claim 8, which is characterized in that

During first procedure of processing, the radium-shine light beam be in order through the first acoustics-optical deflector and Second acoustics-optical deflector the transmitting, and

The first linear frequency-modulated sound wave and the second linear frequency modulation acoustic wave relative to during first processing The point size of the radium-shine light beam at the workpiece and change the radium at the workpiece during second processing The point size of irradiating light beam.

11. a kind of radium-shine system of processing contains the workpiece of the metal layer of dielectric layer and the dielectric layer that overlaps for micro Process, and In the first procedure of processing, radium-shine light beam is guided on the workpiece to cut and penetrate the metal layer, and is added second During work step is rapid, the radium-shine light beam is guided on the workpiece to process the dielectric layer, the radium-shine system of processing packet Contain:

Radium-shine source, to generate radium-shine light beam；And

Acoustics-optical deflector subsystem, it includes:

First acoustics-optical deflector is configured to receive the radium-shine light beam, wherein the first acoustics-optical deflector Be in operation for when by the first radio frequency signals drive, generate the radium-shine light beam for being enough to be inputted in a first direction in The acoustic wave of diffraction is carried out, and thereby exports the first diffraction radium-shine light beam；

Second acoustics-optical deflector is configured to receive the first diffraction radium-shine light beam, wherein the rising tone-light Learning in deflector operation is for when by the second radio frequency signals drive, and generation is enough the first diffraction radium-shine light beam the The acoustic wave of diffraction is carried out in two directions, and thereby exports the second diffraction laser light from the acoustics-optical deflector subsystem Beam, the second direction be not parallel with the first direction or reversed；

Locator is configured to receive the radium-shine light beam exported from the acoustics-optical deflector subsystem, the positioning Device is configured to assign the relative motion of the radium-shine light beam along the radium-shine light beam track relative to workpiece；

Computer is for executing machine readable instructions, to control the first acoustics-optical deflector and described in operation The operation of two acoustics-optical deflector；And

Machine-readable medium can be had the instruction being stored thereon, be worked as institute by the computer to access, the machine-readable medium When stating instruction by computer execution, so that in the first acoustics-optical deflector and the second acoustics-optical deflector Each can generate linear frequency modulation acoustic wave wherein during second processing.

12. radium-shine system of processing as claimed in claim 11, which is characterized in that the machine-readable medium, which has, is stored in it On instruction, when described instruction is executed by the computer, so that the first acoustics-optical deflector and the rising tone - optical deflector generates during first procedure of processing is performed the same path along radium-shine light beam track at it In linear frequency modulation acoustic wave.

13. a kind of for forming the radium-shine system of processing of workpieces processing, the radium-shine system of processing using processing radium-shine light beam Include:

Radium-shine source, to generate the processing radium-shine light beam；

Acoustics-optical deflector subsystem is operated to deflect the processing radium-shine light beam along two axis；

One or more processors are configured to control the operation of the acoustics-optical deflector subsystem, so that the sound It is primary that-optical deflector subsystem at least deflects the processing radium-shine light beam within the time cycle, wherein in week time During phase, the radium-shine light beam is continuous wave radium-shine light beam.

14. radium-shine system of processing as claimed in claim 13, which is characterized in that the radium-shine source is the radium-shine source of continuous wave.

15. radium-shine system of processing as claimed in claim 13, wherein the acoustics-optical deflector subsystem includes:

First acoustics-optical deflector is configured and operationally to deflect the processing radium-shine light beam along first axle；With And

Second acoustics-optical deflector is configured and operationally to deflect the processing radium-shine light beam along the second axis.

16. radium-shine system of processing as claimed in claim 15, which is characterized in that the first axle is perpendicular to described second Axis.

17. radium-shine system of processing as claimed in claim 13 also includes locator, it is used to deflect the processing radium in operation Irradiating light beam is along at least one axis, wherein the acoustics-optical deflector subsystem is configured among path, the processing radium Irradiating light beam can be propagated along position of the path between the radium-shine source and the locator.

18. a kind of method, it includes:

During the time cycle, the radium-shine light beam of continuous wave is generated；

It is more to generate from the radium-shine light beam along two axis that the radium-shine light beam is deflected using acoustics-optical deflector subsystem A laser pulse；And

Carry out workpieces processing using the multiple laser pulse.

19. method as claimed in claim 18 is further included using the radium-shine source of continuous wave and is generated described in continuous wave Radium-shine light beam.